Threat Research Feed

10/29/26
n0n Credential-Driven Destructive Extortion
CRITICAL
+ —
Intel Source:
SOCRadar
Intel Name:
n0n Credential-Driven Destructive Extortion
Date of Scan:
10/29/26
Impact:
CRITICAL
Summary:
The n0n operation combined credential theft, data exfiltration, encryption, and deliberate deletion of enterprise data and backups. SOCRadar recovered an exposed operator workspace containing 1,719 files, including attack scripts and encryptor source code, that implicated 23 organizations across 14 countries. The publisher distinguishes 15 organizations subsequently published on the leak site, four compromised organizations not published, and four targets with access attempts but no demonstrated outcome. Documented entry routes used weak or stolen credentials, including FortiGate credential stuffing, exposed AWS keys, remote administration credentials, and reused hosting sessions; the recovered evidence did not establish vulnerability exploitation. At a sports retailer, an exposed cloud key enabled collection from 40 storage buckets and destructive deletion. Other cases involved theft of hosting backups and manipulation of an ISP's firewall policies. The operator claimed extensive network disruption, but that claim should remain separate from the documented policy changes. External scripts stopped services and removed recovery options before encryption, making recovery depend on backups outside the compromised administration plane. AI tooling helped produce and debug scripts, while the evidence supports a human-directed operation. Prioritize phishing-resistant authentication and revocation of exposed credentials across remote access and management portals. Restrict firewall policy administration and cloud deletion privileges, and investigate unusual token creation and bulk storage operations. Preserve independent, immutable backups and correlate endpoint, cloud, and firewall evidence during response.
Source: https://socradar.io/blog/n0n-ransomware-ai-assisted-extortion/
10/29/26
KeyDrift Intrusions into Cisco SD-WAN Controllers
HIGH
+ —
Intel Source:
Cisco Talos
Intel Name:
KeyDrift Intrusions into Cisco SD-WAN Controllers
Date of Scan:
10/29/26
Impact:
HIGH
Summary:
UAT-11473 exploited CVE-2026-20127 to gain unauthorized access to Cisco Catalyst SD-WAN controllers and installed the KeyDrift framework inside their control-plane environment. Cisco Talos reports targeting of space and satellite services, defense and aerospace technology, and policy research. Its October 28 analysis documents persistent controller access and collection mechanisms, although the activity described is historical and does not establish that every affected deployment remains compromised. The PolyFlame component enters legitimate controller processes through modified service launchers and process injection. It intercepts socket and encrypted control-channel operations, supports file access and command execution, and can relay commands between the operator and other peers. This placement gives the framework access to sensitive state inside trusted network-control processes. LittleFilch separately hooks decrypted control traffic, extracts session identifiers and what Talos assesses to be data-plane keying material, then encrypts collected records for staging or transmission. Talos could not recover its runtime destination configuration, and the exact in-band tasking format remains unresolved. The publisher assesses substantial infrastructure overlap with Volt Typhoon's covert access network with high confidence; the overlap does not resolve attribution of all the activity. Apply Cisco's controller patches and deployment hardening guidance first. Review service launchers, injected modules, and staging artifacts against the report's findings. Activate incident response and engage Cisco support when post-compromise evidence appears, because patching alone does not establish controller integrity.
Source: https://blog.talosintelligence.com/uat-11473-keydrift/
10/29/26
Doorway Botnet Compromises Network Devices and Java Servers
HIGH
+ —
Intel Source:
CloudSEK
Intel Name:
Doorway Botnet Compromises Network Devices and Java Servers
Date of Scan:
10/29/26
Impact:
HIGH
Summary:
The Doorway operation compromised routers, cameras, and exposed enterprise services to build a remotely accessible device fleet and proxy infrastructure. CloudSEK identified two identical exposed workspaces in October containing approximately 41 GB of operational material, including a Rust exploitation engine, 59 exploit modules, credential records, and execution evidence. The publisher observed the operation as active during its investigation. Evidence includes 7,369 successful RouterOS authentications across 4,923 device identities, 20 Jenkins hosts with confirmed remote code execution, and 16 Openfire plugin uploads. These measures represent different outcomes and should not be summed as unique compromised systems. An operator ledger reports substantially larger join and download totals, but downloads and repeated join events do not independently prove successful execution. The framework fingerprints services before selecting attacks, reuses default or exposed credentials, and supports multiple delivery methods and processor architectures. Payload retrieval can be signed and bound to the target address, while beacon evidence helps distinguish code execution from file delivery. Recovered components include a reverse shell and server-specific loaders. The final clients' full command protocol, some RouterOS exploitation outcomes, and commercial monetization remain unconfirmed. First restrict exposed management interfaces, rotate shared device credentials, and patch or isolate affected Java services. Then use the published indicators and detection rules to investigate delivery traffic, beacons, unexpected proxy settings, and unauthorized plugins. Preserve configuration history and network logs, and classify confirmed execution separately from probes and downloads.
Source: https://www.cloudsek.com/blog/doorway-botnet-global-iot-router-camera-enterprise-server-attacks
10/29/26
LiteLLM 1.107.0 Supply-Chain Credential Theft
HIGH
+ —
Intel Source:
ThreatMon
Intel Name:
LiteLLM 1.107.0 Supply-Chain Credential Theft
Date of Scan:
10/29/26
Impact:
HIGH
Summary:
The compromised LiteLLM 1.107.0 release placed credential-stealing startup hooks into Python environments reached through official package and container distribution. ThreatMon's October 28 research describes its inspection of the trojanized wheels and explains why exposure extends beyond applications that explicitly import LiteLLM. Installed startup files sit at the environment level, so an ordinary Python invocation can launch collection on a developer workstation, build runner, test container, or gateway host. The report traces release-pipeline takeover to stolen maintainer credentials and a workflow redirected to a lookalike repository, with a second compromised account used for deployment approval. The collector reads developer and AI-tool authentication data, cloud credentials, browser stores, private keys, process environments, and accessible application database fields. It encrypts collected data and attempts upload to an attacker-controlled service. Startup execution alone does not establish successful collection or exfiltration; response teams must assess each environment. A separate Unix loader can retrieve and execute additional Python code, but ThreatMon had not observed that component serving payloads in the wild. The malicious release was quarantined and affected container images removed, which does not remediate environments that already installed them. Identify and isolate systems that obtained the release, using inspection methods that do not start Python inside a suspect environment. Rebuild affected environments from trusted images and terminate surviving malicious processes. Rotate accessible credentials and audit their subsequent use across cloud, source-control, registry, and application services.
Source: https://threatmon.io/litellm-1-107-0-from-startup-hooks-to-credential-theft/
10/29/26
Help Desk Vishing Targets Canadian Enterprise Identities
HIGH
+ —
Intel Source:
Canadian Centre for Cyber Security
Intel Name:
Help Desk Vishing Targets Canadian Enterprise Identities
Date of Scan:
10/29/26
Impact:
HIGH
Summary:
Help desk impersonators are targeting Canadian organizations with voice phishing that relays authentication through fraudulent sign-in pages and can yield authenticated enterprise sessions. The Canadian Centre for Cyber Security's October 28 alert identifies targeting across multiple firms, with finance and legal organizations increasingly targeted since July and October activity concentrated on the legal sector. Callers claim that employees must urgently migrate passkeys, enroll authentication methods, or change passwords. They contact personal or office phones and collaboration services, sometimes spoofing support numbers or referencing internal personnel. Organization-branded phishing pages relay credentials and multi-factor responses to legitimate identity providers, allowing conventional multi-factor authentication to be bypassed through real-time interception. The alert includes concrete campaign-associated domains and notes frequent infrastructure rotation. Reported post-compromise behavior includes registration of attacker-controlled authentication methods, deletion of security notifications, password resets for applications outside single sign-on, and automated access to Microsoft 365 and connected services. Data theft can lead to extortion. The Cyber Centre attributes parts of that post-compromise account to cited reporting; it does not publish a victim count or establish that every targeted Canadian firm suffered those outcomes. Prioritize phishing-resistant authentication, managed-device access policies, and independent verification of unexpected support calls. Monitor authentication-method changes, security-mail deletion, and unusual bulk file access. If compromise is suspected, revoke sessions and tokens, remove unauthorized methods, reset affected credentials, and preserve identity and Microsoft 365 audit logs.
Source: https://www.cyber.gc.ca/en/alerts-advisories/al26-028-ongoing-help-desk-voice-phishing-campaign-targeting-canadian-organizations
10/29/26
Node2Shell Exploitation Attempts Against Brotli Decoders
MEDIUM-HIGH
+ —
Intel Source:
Zscaler
Intel Name:
Node2Shell Exploitation Attempts Against Brotli Decoders
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
Attackers probed and attempted to exploit vulnerable Brotli decoders shortly after disclosure of CVE-2026-111497, also called Node2Shell or HTTP2Shell. Zscaler ThreatLabz recorded 402 events from 20 source addresses targeting decoys associated with 12 customers during the first 24 hours. The report classifies 40% as reconnaissance and 60% as exploitation attempts, and provides activity-linked network indicators and fingerprints. These observations document hostile activity against exposed decoys. Zscaler does not confirm compromise of customer production systems. The vulnerability affects block-decoding logic and can produce an out-of-bounds write when malformed compressed data reaches a vulnerable implementation. Potential consequences include crashes, memory disclosure, and remote code execution, with actual exposure depending on application behavior and the bundled decoder. Affected paths include servers decompressing incoming requests and clients processing responses from external or user-controlled services. Node runtimes, Python bindings, service proxies, and other products can carry separate vulnerable copies, so a host-level library update may leave workloads exposed. The report references code execution demonstrated elsewhere in a test environment; that result should not be presented as an observed production intrusion. Prioritize updates for components that decompress untrusted data and rebuild affected containers with patched dependencies. Where updates are delayed, disable vulnerable decoding paths where supported and tested. Investigate malformed compressed traffic, correlated worker crashes, unexpected child processes, and subsequent outbound connections using the report's activity evidence.
Source: https://www.zscaler.com/blogs/security-research/cve-2026-111497-node2shell-http2shell-vulnerability-may-allow-rce
10/29/26
INJ3CTOR3 FreePBX Campaign Adds Credential Exfiltration
MEDIUM-HIGH
+ —
Intel Source:
Cyble
Intel Name:
INJ3CTOR3 FreePBX Campaign Adds Credential Exfiltration
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
INJ3CTOR3's JOMANGY campaign now includes repeated collection of telephone-system credentials, extending toll-fraud exposure beyond the infected PBX. Cyble compared earlier campaign artifacts with material collected from two operator staging hosts in September 2026. The updated installer contains commands to upload 17 configuration files covering SIP trunks, extensions, management access, databases, and mail relays. Recurring scheduled execution can repeat that collection every minute. Cyble observed the upload implementation in recovered code, rather than capturing an actual transfer, and it holds no victim list for this update. The report does not measure how many credentials reached the operator. The existing self-repairing persistence remains largely intact, while watermarks, access-gate values, and staging identifiers have changed. Encoded requests and rewritten web-server rules can reach a backdoor without naming it in access logs. Lab validation also showed why the management console can miss configuration written directly into the underlying telephone service. The staging servers stopped responding in late September, but their disappearance does not revoke previously exposed secrets. Isolate suspected systems and preserve scheduled tasks, account records, configurations, and carrier logs before remediation. Rebuild from a clean image or remove persistence from the system layer upward, then rotate every affected secret after the host is clean. Coordinate carrier restrictions, registration monitoring, destination controls, and spending limits to contain fraud using stolen trunk credentials.
Source: https://cyble.com/blog/hijacked-pbx-to-stolen-trunk-inj3ctor3s-campaign/
10/29/26
Fake DocuSign Installer Establishes Hexnode Persistence
MEDIUM-HIGH
+ —
Intel Source:
Huntress
Intel Name:
Fake DocuSign Installer Establishes Hexnode Persistence
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
A phishing-delivered Hexnode management agent allowed an attacker to reinstall unauthorized ScreenConnect after the initial remote-access client was removed. Huntress investigated two alerts on the same endpoint, separated by several hours, that connected to the same attacker-controlled relay. The incident shows why removing the visible remote-control tool can leave the underlying management access intact. Proposal-themed emails referenced an organization with which the target likely had a business relationship and directed the recipient to an installer disguised as a DocuSign-related application. The package actually installed legitimate Hexnode management components, including its gateway and agent, with supporting database software. The attacker then used that management access to deploy ScreenConnect. After the service provider removed the first installation, another copy appeared and reconnected to the same relay. Huntress isolated the endpoint following the second alert and began broader remediation. The report describes abuse of legitimate administration software through social engineering; it does not identify a vulnerability in Hexnode or establish the extent of data access. It provides delivery and relay indicators, but no confirmed victim geography or campaign-wide count. First investigate unexpected management agents and enrollments whenever unauthorized remote access reappears, correlating installation history with network connections. Contain the endpoint and remove unauthorized management enrollment and attacker-created access, then revoke exposed sessions and credentials. Maintain an inventory of approved management tenants and remote-access instances, and reinforce verification of proposal and document-signing requests.
Source: https://www.huntress.com/blog/attackers-turn-mdm-into-remote-access
10/29/26
Fake Teams Invitation Enables Rippling-Delivered Remote Access
MEDIUM-HIGH
+ —
Intel Source:
Huntress
Intel Name:
Fake Teams Invitation Enables Rippling-Delivered Remote Access
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
An attacker used a tax-planning email and a lookalike Microsoft Teams invitation to establish unauthorized Rippling management access, then deploy ScreenConnect with SYSTEM privileges. Huntress traced the incident from a receptionist's click through endpoint enrollment evidence and remote-access installation. The linked page presented itself as a calendar and redirected through another page that appears to have facilitated enrollment into an attacker-controlled management tenant. Because the site was taken down shortly afterward, the publisher describes parts of that enrollment mechanism as an assessment. About a minute after the click, an enrollment-related registry artifact appeared. The management agent was subsequently downloaded and installed with SYSTEM permissions. Approximately 35 minutes later, the attacker used the management platform to stage and run a ScreenConnect installer that connected to an unauthorized relay and triggered an alert. Huntress determined that the activity originated from the online management agent, suggesting an account had been created to support the attack. The report does not establish a wider victim count, affected geography, or subsequent data theft. Its operational finding is a privileged deployment path created by social engineering and legitimate management functionality. Prioritize correlation of unexpected invitations, device enrollment, agent installation, and remote-access connections across endpoint and management telemetry. Isolate suspected endpoints, remove unauthorized enrollment, and disable attacker-created access before treating remote-tool removal as complete. Review approved management tenants and applications, revoke affected sessions and credentials, and reimage systems whose integrity cannot be established.
Source: https://www.huntress.com/blog/attackers-turn-mdm-into-remote-access
10/29/26
CyberDakait Android Banking Fraud and SMS Theft
MEDIUM-HIGH
+ —
Intel Source:
Group-IB
Intel Name:
CyberDakait Android Banking Fraud and SMS Theft
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
The CyberDakait banking-fraud campaign uses coordinated calls and messages to move Indian bank customers from a fraudulent protection-plan cancellation process into Android malware installation. Group-IB identified 17,960 compromised devices across exposed collection channels and 7,611,443 intercepted messages, including authentication codes and transaction notifications. Those are aggregate campaign findings and do not establish an equal distribution across its seven malware families. The publisher observed impersonation of 11 banking institutions and two government programs. Callers claim that a paid protection plan must be deactivated, then send a phishing link that collects card details, banking credentials, and identity information while initiating a malicious application download. GitHub-hosted payloads and fake application-update screens support delivery of additional components, while battery-optimization prompts help maintain background execution. Some analyzed chains also deploy a cryptocurrency miner. Group-IB distinguishes seven families by their communication protocols, database structures, and message formats; shared packaging and themes do not prove that every family has one developer or operator. The report also documents an administrative application used to view stolen information and issue commands. Identified distribution accounts and repositories were reported to GitHub and deactivated. Banks should prioritize customer verification of protection-plan calls and rapid response to reported application installation or fraud. Use the published indicators to investigate distribution and mobile collection infrastructure, and report replacement accounts and repositories. Restrict untrusted application installation, review sensitive permissions, and promptly freeze affected banking access when compromise is suspected.
Source: https://www.group-ib.com/blog/cyberdakait-seven-android-trojans-indian-banking/
10/29/26
MR ROBOT Banking Malware-as-a-Service Operation
MEDIUM-HIGH
+ —
Intel Source:
CloudSEK
Intel Name:
MR ROBOT Banking Malware-as-a-Service Operation
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
MR ROBOT rents Android banking trojans through an affiliate platform that supports customized financial-service lures, stolen-message collection, and remote management of infected devices. CloudSEK's investigation links a Telegram-based builder, an operator console, and a versioned web panel to a catalog of approximately 36 impersonated banking, utility, and government-service applications. One tracked affiliate showed approximately 154 enrolled devices and 3,270 intercepted messages in its panel. A separate exposed tenant contained 153 devices and 5,407 messages, with call forwarding active on 58% of those victims. The observations describe different backends and should not be treated as a complete platform-wide victim count. Sideloaded droppers decrypt a second-stage application that intercepts messages and authentication codes, redirects incoming calls, and displays phishing forms for banking and identity information. Collected data reaches cloud-hosted databases and Telegram. Per-build signing identities, delayed activation, missing launcher icons, and local network interference complicate detection. CloudSEK identifies similarities to previously reported banking malware services but does not establish common ownership. Prioritize bank fraud investigations when customers report missed verification calls or unexpected identity-update applications, and move away from message-based authentication where feasible. Hunt the published application and infrastructure indicators, review message and call permissions, and coordinate malicious-backend takedown. Reinforce use of official application channels and verification of unsolicited banking-update requests.
Source: https://www.cloudsek.com/blog/inside-mr-robot-malware-as-a-service-targeting-indian-bank-customers
10/29/26
WhatsApp Finance Lures Deliver Overlord Remote Access
MEDIUM-HIGH
+ —
Intel Source:
NetbyteSEC
Intel Name:
WhatsApp Finance Lures Deliver Overlord Remote Access
Date of Scan:
10/29/26
Impact:
MEDIUM-HIGH
Summary:
A WhatsApp account-statement lure delivered a Windows loader and the Overlord remote-access agent to finance and business users. NetbyteSEC ties the analyzed delivery to October 12-13 and infers Malaysian targeting from the language and context of one sample; the report does not measure the campaign-wide victim population. The archive pairs a legitimate Microsoft-signed executable with an unsigned malicious library. Running the apparently trusted program loads the adjacent attacker library, which resolves operating-system functions dynamically and uses direct system calls to reduce visibility through hooked interfaces. The loader retrieves an encrypted second-stage container from hosted release infrastructure, with alternative delivery servers, and retains an encrypted local copy for later execution. It starts a legitimate Windows process suspended, replaces its executable image, and resumes execution inside that process. The resulting Go-based agent communicates over an encrypted WebSocket channel. Analysis identifies remote desktop control, keylogging, file operations, shell execution, proxying, and plugin-loading capabilities; those capabilities do not prove that operators exercised every function against victims. The publisher distinguishes the Overlord client from TankRat-branded panel infrastructure and leaves actor attribution unresolved. Prioritize review of finance-themed WhatsApp attachments and execution of signed programs from newly extracted directories. Hunt the report's loader, cached-payload, persistence, and network artifacts, correlating them with process-injection evidence. Investigate exposed credentials and sensitive activity on affected endpoints, while separating confirmed execution from the agent's broader analyzed capabilities.
Source: https://notes.netbytesec.com/2026/10/malaysian-whatsapp-phishing-wave.html
10/29/26
Fake DeepSeek Downloads Deliver Kong RAT
MEDIUM
+ —
Intel Source:
Hybrid Analysis
Intel Name:
Fake DeepSeek Downloads Deliver Kong RAT
Date of Scan:
10/29/26
Impact:
MEDIUM
Summary:
An SEO-poisoning campaign places fake DeepSeek download sites in search results and uses large installer archives to deliver Kong RAT. Hybrid Analysis describes malicious lookalike sites offering a supposed local deployment tool, with a signed application lending apparent legitimacy to a package that also drops an attacker downloader. The report connects the delivery pattern to earlier targeting of developer utilities, while treating continuity as an assessment. It does not provide a confirmed victim count or affected-country distribution. The downloader retrieves a file presented as video content that actually contains shellcode and a library. That stage creates a local program directory, reports staging information, and retrieves a legitimate launcher alongside an attacker library and encrypted payload data. Scheduled execution initiates library side-loading. Additional persistence changes the Windows logon configuration, allowing the malicious chain to start again with user sessions. The resulting Kong RAT includes keyboard and clipboard collection, stores captured information locally, and communicates with the campaign's infrastructure. The publisher's suggestion of Chinese actor involvement rests on development-path clues and should not be treated as established attribution or victim geography. Prioritize validation of DeepSeek installers and developer-tool downloads against approved distribution channels. Hunt the published delivery indicators and endpoint artifacts, correlating scheduled execution, side-loaded libraries, and unexpected logon changes. Investigate keystroke and clipboard exposure on affected systems and remove the documented persistence alongside the delivered components.
Source: https://hybrid-analysis.blogspot.com/2026/10/seo-poisoning-campaign-targets-deepseek.html
10/22/26
PHP2Shell exploitation and updated AWS WAF protections
CRITICAL
+ —
Intel Source:
Amazon Web Services
Intel Name:
PHP2Shell exploitation and updated AWS WAF protections
Date of Scan:
10/22/26
Impact:
CRITICAL
Summary:
Attackers sent PHP2Shell exploitation attempts to Amazon's MadPot sensors before public disclosure, making internet-facing PHP workloads a priority for patching and retrospective investigation. AWS's October 22 update adds version 2.3 of its PHP application managed rule group, with new protection for CVE-2026-109619. The directly linked AWS investigation records malicious requests beginning October 17 and additional attempts before the October 21 patches. Sensor observations establish real attack traffic, but do not establish how many production systems were compromised. The two PHP engine vulnerabilities can leave stale request state after memory exhaustion and enable remote code execution before application authentication provides protection. Their behavior depends on the runtime and deployment configuration, so application-level controls cannot substitute for fixing PHP. AWS supplies usable custom WAF logic and explains where the new managed protection stops: it addresses CVE-2026-109619, while request-body size controls provide partial mitigation for CVE-2026-109618. Body inspection limits and allowed upload paths can leave gaps, and uninspected oversized content requires separate handling. Deploy the fixed PHP releases, including versions 8.2.35, 8.3.36, 8.4.27, and 8.5.12, and investigate exposed workloads against the reported attack timeline. Apply the updated managed protection and complementary body-size controls, briefly evaluate logged matches before blocking, and keep upload exceptions narrow. Review runtime mitigations alongside edge controls, recognizing that WAF coverage reduces exposure while the PHP upgrade supplies the complete fix.
Source: https://aws.amazon.com/blogs/security/using-aws-waf-to-apply-mitigations-for-php2shell-rce-vulnerabilities-cve-2026-109618-and-cve-2026-109619/
10/22/26
Microsoft 365 token replay and account persistence campaign
HIGH
+ —
Intel Source:
Huntress
Intel Name:
Microsoft 365 token replay and account persistence campaign
Date of Scan:
10/22/26
Impact:
HIGH
Summary:
Attackers used stolen Microsoft 365 sessions to compromise business email accounts at 79 organizations between July 14 and September 22, 2026. Huntress recorded at least 497 successful sign-ins alongside more than 200,000 failed attempts. Its October 22 report shows how a session theft campaign became durable identity compromise and business email abuse. Adversary-in-the-middle relays captured tokens after victims completed MFA; device-code phishing and social engineering supplied additional sessions. Scripted and browser-based replay then reached Microsoft Graph, Exchange Online, SharePoint Online, and device registration services. The operator registered at least 48 rogue Entra devices. At 19 organizations, existing MFA methods were cleared and replaced within three hours of the first campaign sign-in. Those changes created persistence that a password reset alone would leave intact. Hidden inbox rules concealed communications, and one compromised account sent a payroll-related message to an outside contact. The report does not establish resulting financial losses or the geography of affected organizations. Infrastructure used for session capture differed from infrastructure used for replay and mailbox actions, limiting the value of blocking one host. Contain suspected accounts by revoking sessions and resetting passwords, then remove unauthorized devices, MFA methods, and mailbox rules. Apply Conditional Access across users, cloud applications, and client types; restrict device-code flows and require approval for new application consent. Prioritize investigations using evidence of valid credentials and successful access, rather than failed-attempt volume alone, and correlate sign-ins with subsequent identity and mailbox changes.
Source: https://www.huntress.com/blog/token-replay-spray
10/22/26
MythicRatatoskr intrusion preceding destructive activity in Ukraine
HIGH
+ —
Intel Source:
ESET
Intel Name:
MythicRatatoskr intrusion preceding destructive activity in Ukraine
Date of Scan:
10/22/26
Impact:
HIGH
Summary:
A Ukrainian grain logistics company suffered an intrusion that progressed from spearphishing to persistent remote access and later destructive activity. ESET's October 22 research reconstructs the 2025 compromise and introduces technical findings on MythicRatatoskr, a Go backdoor that communicates through the Yggdrasil overlay network. The report attributes initial access to UAC-0099: a malicious attachment on July 16 deployed the MATCHBOIL and MATCHWOK toolchain, followed by MythicRatatoskr on July 17. Additional systems received tools through OpenSSH file transfers. The backdoor established Windows service persistence and supported host profiling, screenshots, file transfer, and proxy access. Yggdrasil supplied encrypted IPv6 connectivity, complicating conventional inspection of direct command-and-control traffic. The intrusion also included separate SOCKS proxy components; ESET could not fully explain why both proxy mechanisms were present. On September 17, Sandworm used Active Directory Group Policy to deploy SDelete for destructive file wiping. ESET assesses a handoff between UAC-0099 and Sandworm with high confidence, while leaving unresolved which actor operated MythicRatatoskr during the intervening period. The observed scope is one company; the report does not establish wider deployment or current activity. Hunt for the published backdoor and proxy artifacts together with unexpected Windows services, OpenSSH transfers, and overlay-network communications. Examine Group Policy changes and administrative execution for destructive tool deployment. Scope response across the full phishing, persistence, and lateral-access sequence so removal of one implant does not leave the earlier foothold available.
Source: https://www.welivesecurity.com/en/eset-research/mythicratatoskr-yggdrasil-enabled-backdoor-deployed-uac0099-sandworm-collaboration/
10/22/26
BlackX UEFI bootkit targeting Central Asian diplomatic agencies
HIGH
+ —
Intel Source:
Kaspersky
Intel Name:
BlackX UEFI bootkit targeting Central Asian diplomatic agencies
Date of Scan:
10/22/26
Impact:
HIGH
Summary:
Attackers deployed the BlackX UEFI bootkit in diplomatic agency networks in Tajikistan, Uzbekistan, and Turkmenistan during early 2026. Kaspersky's October 22 report documents an espionage operation in which existing network compromise and administrator privileges preceded bootkit installation. BlackX exploited Baton Drop, CVE-2022-21894, to install custom Machine Owner Keys and load attacker-controlled EFI components. Its implementation resembles BlackLotus, although Kaspersky could not determine whether the operator reverse-engineered that tool or acquired its source code. A kernel driver protected bootkit files and attempted to interfere with security callbacks and event tracing. Some routines specifically targeted Kaspersky or Avast products; Kaspersky's dynamic analysis found that the relevant routines did not interfere with its products. Two user-mode payloads were observed in the wild. One used Google Drive, OneDrive, or Yandex cloud storage for command exchange and further payload loading. The other, OxiDLLRunner, used DLL sideloading and an encrypted Rust payload to obtain additional plugins. Kaspersky attributes the operation to APT41 with medium confidence, based on overlapping implementation and tooling evidence. Fully patched systems remained exposed where the additional Secure Boot certificate revocations had not been activated, demonstrating that installed Windows updates alone did not establish protection. Verify activation of the required Secure Boot revocations using the vendor's deployment procedure. Investigate boot-partition changes, malicious driver activity, and sideloaded payloads together, then correlate endpoint findings with unusual cloud-storage command traffic and investigate the administrator-level compromise that enabled installation.
Source: https://securelist.com/uefi-bootkit-blackx/121539/
10/22/26
ARTEX-assisted attacks against software and IT vendors
MEDIUM-HIGH
+ —
Intel Source:
Wiz
Intel Name:
ARTEX-assisted attacks against software and IT vendors
Date of Scan:
10/22/26
Impact:
MEDIUM-HIGH
Summary:
Attackers used ARTEX tooling against multiple IT and software vendors in October 2026, probing identity integrations and tenant boundaries that could provide access beyond a single account. Wiz's October 22 investigation documents activity overlapping tooling associated with Storm-0940, while treating shared tooling as evidence to assess rather than proof of identical operators. The observed sessions combined application reconnaissance, account abuse, and automated credential attacks. Operators retrieved JavaScript source maps, configuration material, and API documentation to identify internal routes and authentication behavior. They tested OAuth callbacks, state handling, unsigned tokens, account recovery, and access across tenant boundaries. Disposable accounts supported these experiments. Coordinated password sprays targeted hundreds of engineers, with request timing and infrastructure rotation consistent with automation. Some sessions extended across 24 to 30 hours; shorter attack bursts and rapid changes in source infrastructure complicated simple rate-based monitoring. The technical detail supports concern about software-provider identity and supply-chain exposure, but most observed attempts failed. The report does not establish that every target was compromised or that customer information was stolen. Geography for the affected vendors is not specified. Correlate application and identity telemetry to connect reconnaissance, coordinated sign-in failures, account-recovery activity, and later successful access. Investigate repeated cross-tenant authorization failures and changes in attack source during a session. Review exposed source maps, configuration, and API documentation for information that assists reconnaissance, and test SSO validation and tenant isolation against the specific abuse paths described by Wiz.
Source: https://www.wiz.io/blog/artex-in-the-wild-storm-0940-expands-targeting-of-it-and-software-vendors
10/22/26
Global ARTEX-assisted attacks seeking personal and financial data
MEDIUM-HIGH
+ —
Intel Source:
Zscaler
Intel Name:
Global ARTEX-assisted attacks seeking personal and financial data
Date of Scan:
10/22/26
Impact:
MEDIUM-HIGH
Summary:
An operator combined ARTEX, AI coding assistants, and custom automation to attack organizations across multiple countries and sectors, with an apparent focus on monetizing personal and transaction data. Zscaler's October 21 investigation expands earlier reporting on South Korean financial targets using recovered infrastructure, configurations, and prompt histories spanning August 4 through October 4, 2026. The records show targeting of retail, education, logistics, dating, SMS, cryptocurrency, payment, and legal-services platforms. Requests sought administrative access, customer records, payment credentials, and datasets containing telephone numbers. Questions about underground data markets support Zscaler's assessment of likely financial motivation, but the report does not provide a verified total of successful compromises or stolen records. The operator customized ARTEX with web, network, and Android analysis tools, divided work among AI sessions, and directed models to review one another's findings. Shared project records and synchronization scripts preserved progress across accounts and servers. A jump server, project-specific proxies, and maintained tunnels separated the working environment from target-facing traffic; cloud archives retained reports, evidence, and data exports. Configurations named several model families, although Zscaler could not independently verify which underlying models served every session. Reduce public exposure and enforce least-privilege access to applications and data stores. Investigate connections to the documented operator infrastructure alongside unusual bulk queries, exports, and access to application secrets. Prioritize exposed attack surfaces and use controlled deception systems to identify unauthorized exploration.
Source: https://www.zscaler.com/blogs/security-research/inside-global-ai-assisted-hacking-campaign
10/22/26
RFRx003 Linux backdoor in an IT company intrusion
MEDIUM-HIGH
+ —
Intel Source:
Solar
Intel Name:
RFRx003 Linux backdoor in an IT company intrusion
Date of Scan:
10/22/26
Impact:
MEDIUM-HIGH
Summary:
Attackers stole mail and maintained privileged Linux access during an intrusion into an IT-sector organization investigated by Solar 4RAYS in summer 2026. The October 22 report introduces RFRx003, a previously undocumented backdoor recovered during the investigation. The foothold dated back several months and involved exploitation of a recently disclosed remote-code-execution vulnerability in an unnamed perimeter service, followed by privilege escalation. The operator also used valid employee credentials against exposed web and VPN services, avoiding a password-guessing pattern. Investigators found suspicious processes and an unauthorized SSH key while reconstructing lateral movement and mailbox theft. RFRx003 supported 16 commands, including system discovery, file operations, interactive terminal access, and SOCKS tunneling. It used HTTPS and secure WebSockets, with Mongoose and mbedTLS components supporting its network communications. Persistence combined a service disguised as time synchronization with a root cron entry. A further payload copy was presented as package metadata, and execution from a memory-backed filesystem reduced reliance on ordinary disk artifacts. The operator also retained a privileged shell for later access and suppressed command history. Solar does not identify the victim's country, the exploited CVE, or a responsible threat group. Audit Linux services, root scheduled jobs, privileged shells, and SSH authorization changes against the report's host artifacts. Correlate abnormal VPN and web logins with mailbox exports, proxy behavior, and unexpected encrypted outbound traffic, preserving volatile evidence where possible. Address the exploited perimeter service and the compromised credentials as part of containment, then check for redundant persistence before closing the incident.
Source: https://rt-solar.ru/solar-4rays/blog/7122/
10/22/26
Armorsteal distribution through poisoned software search results
MEDIUM-HIGH
+ —
Intel Source:
Zscaler
Intel Name:
Armorsteal distribution through poisoned software search results
Date of Scan:
10/22/26
Impact:
MEDIUM-HIGH
Summary:
Poisoned search results directed Windows users to fake Chocolatey and Microsoft Activation Scripts websites that distributed Armorsteal. Zscaler identified the campaign in September 2026 and published its technical investigation on October 22. The lures targeted people seeking software installation commands, creating a plausible route onto developer and administrator workstations. A displayed PowerShell command downloaded the malicious installer; the Chocolatey variant also ran the legitimate installer, allowing the expected software to work and concealing the additional compromise. Both variants requested elevated privileges and installed a Python 3.11 runtime with Pyarmor-protected modules. Scheduled tasks launched the remote-access component at logon and system startup. That component profiled the host, downloaded surveillance and information-stealing modules, and verified RSA-2048 signatures before executing remotely supplied Python tasks. The surveillance module logged keystrokes, monitored clipboard contents, and captured screenshots. The stealer targeted browser cookies, saved passwords, Discord tokens, cryptocurrency material, SSH keys, and other application data. These are documented capabilities; the report does not quantify stolen enterprise data or affected organizations. The fake sites became unavailable in late September and early October, while the operator changed infrastructure during analysis, so those observations do not establish continuing distribution through the same pages. Investigate suspicious PowerShell downloads, new scheduled tasks, and unexpected Python installations using the report's artifacts. Review browser remote-debugging activity and remove unauthorized Microsoft Defender exclusions. Block verified malicious infrastructure and require users to validate software websites and installation commands before execution.
Source: https://www.zscaler.com/blogs/security-research/seo-poisoning-campaign-delivers-new-armorsteal-malware
10/22/26
SiriusQT remote-access and credential-theft campaigns
MEDIUM-HIGH
+ —
Intel Source:
Group-IB
Intel Name:
SiriusQT remote-access and credential-theft campaigns
Date of Scan:
10/22/26
Impact:
MEDIUM-HIGH
Summary:
Multiple operators distributed SiriusQT in malicious Windows campaigns, combining remote control with credential theft and browser-session abuse. Group-IB's October 22 investigation connects the malware's commercial ecosystem, public builds, delivery activity, and technical implementation. Campaigns targeted the United Kingdom, Germany, France, Spain, Italy, and the Netherlands, among other countries. Researchers observed SiriusQT as a final-stage payload delivered through UpCrypter in May 2026 and tracked more than 280 publicly available client builds across different campaigns. Those builds are not a count of victims: the underlying submission data includes real compromises, operator testing, and sandbox activity. License analysis identified at least 30 paying users alongside extensive use of leaked or generated licenses, supporting a distributed operator ecosystem rather than one centrally run campaign. SiriusQT provides hidden virtual network computing, browser automation, keylogging, clipboard monitoring, wallet collection, and credential theft. It can clone browser profiles, preserve access to stored sessions, inject into processes, establish persistence, and use custom encrypted command traffic. These capabilities create exposure for business accounts accessed from infected endpoints, but the report does not quantify resulting enterprise losses. Attribution to a single developer remains uncertain, despite an identifiable reseller and active underground promotion. Restrict untrusted downloads and scrutinize unsolicited files, cracked software, and unexpected installation requests. Hunt for the published client and network artifacts alongside anomalous process injection, browser execution, and outbound communications. Update browsers and endpoint protections, enforce MFA, and investigate suspicious sessions even when the user has not reported an unfamiliar sign-in prompt.
Source: https://www.group-ib.com/blog/siriusqt-rat-hvnc-stealer/
10/22/26
RatHat Android compromise revealed through incident audit logs
MEDIUM-HIGH
+ —
Intel Source:
Zimperium
Intel Name:
RatHat Android compromise revealed through incident audit logs
Date of Scan:
10/22/26
Impact:
MEDIUM-HIGH
Summary:
RatHat established Android control beyond the lifecycle of its visible application, according to infected-device telemetry analyzed in Zimperium's October 22 report. The new findings reconstruct execution, network activity, and persistence using operating-system audit data. The publisher's directly linked original investigation describes malicious distribution through targeted smishing, malvertising, and third-party download portals. A streaming application lure encouraged installation and accessibility permissions. RatHat then automated settings changes, enabled wireless debugging, collected the local pairing code, and paired with Android Debug Bridge without requiring a separate attacker-controlled computer. This gave its native agent shell-level execution. Go components and reverse-proxy processes could continue independently of the original application, making an ordinary uninstall an incomplete response. The incident logs also showed changes to background restrictions, local interprocess communications, and recurring command-server connections approximately every 60 to 90 seconds. The broader toolchain supports remote interaction, capture of sensitive screen input, and interception of one-time codes, creating risk for work accounts accessed on personal or managed phones. The report does not establish victim numbers, affected geography, or confirmed enterprise account losses. Enable supported Android security and network audit collection through device management, and assess the newer intrusion-logging facility where available before an incident occurs. Correlate suspicious installation and accessibility events with debugging activation, native processes, and repeated outbound connections. During containment, investigate shell-level agents and reverse tunnels beyond the application package, and block verified distribution and command infrastructure through mobile threat controls.
Source: https://zimperium.com/blog/beyond-the-sandbox-leveraging-android-intrusion-logs-to-expose-advanced-threats
10/22/26
LLMjacking attempt delivering an obfuscated XMRig miner
MEDIUM
+ —
Intel Source:
Broadcom
Intel Name:
LLMjacking attempt delivering an obfuscated XMRig miner
Date of Scan:
10/22/26
Impact:
MEDIUM
Summary:
An attacker sent an obfuscated cryptocurrency-mining payload to an internet-facing AI honeypot. Broadcom's October 21 Skynet research documents the captured request. The request asked the model to decode a string without explaining its contents. That string carried a compressed shell script intended for direct execution, concealing attack commands from prompt filters. The script downloaded XMRig and could compile it, selected among four mining pools, and concealed the miner under a logging-themed filename. Its directory searches and configuration handling suggest targeting of camera-streaming systems using go2rtc. It also attempted to make a configuration file immutable and repeatedly erase local service access logs. The honeypot captured the payload without executing it; the report does not establish successful mining, downstream compromise, or financial loss. The separate Skynet agent instructions captured the previous day describe broader autonomous exploitation capabilities, but do not prove those capabilities ran against victims or establish a common operator. Require authentication on every AI API path and restrict tool-enabled agents' outbound access and privileges. Rotate and scope AI keys while monitoring inference use and spending. Investigate matching payload and network artifacts, including suspicious miner execution and service-log deletion, without treating receipt of a malicious prompt as proof of successful compromise.
Source: https://www.security.com/threat-intelligence/ai-llmjacking-skynet
10/21/26
PHP2Shell: Active Exploitation Attempts and Enterprise PHP Exposure
CRITICAL
+ —
Intel Source:
Wiz; CrowdStrike; Imperva; Armis; Cloudflare; The PHP Group
Intel Name:
PHP2Shell: Active Exploitation Attempts and Enterprise PHP Exposure
Date of Scan:
10/21/26
Impact:
CRITICAL
Summary:
Exploitation attempts against PHP2Shell accompanied disclosure of two critical PHP runtime vulnerabilities, while research across multiple vendors demonstrated unauthenticated code execution and sensitive-memory disclosure. Tracked as CVE-2026-109618 and CVE-2026-109619, the flaws undermine request processing inside PHP itself, placing enterprise websites, shared hosting, containerized applications, and PHP-based management interfaces at risk even when application code is secure. Cloudflare observed attack traffic and changing payloads intended to bypass protections. Imperva reported no matching traffic during its pre-disclosure monitoring; the reviewed reports do not quantify successful compromises or confirmed downstream theft. The first vulnerability can leave request data incompletely initialized after memory exhaustion, allowing cleanup to consume invalid state. The second can retain pointers to released request memory that a later request on the same worker accesses. Demonstrated consequences include code execution under the web worker's privileges, process-memory disclosure, arbitrary file reads, and worker crashes. Exposed secrets can include database credentials, API keys, and authentication material. Apache mod_php presents the clearest demonstrated risk before application code executes. PHP-FPM and other deployments also require assessment, but version, enabled features, request limits, and worker behavior affect exploitability. Affected code paths extend back as far as PHP 5.4; this does not establish identical exposure across every historical release. Normal HTTP responses and sparse application logs do not rule out exploitation. First, inventory internet-facing and embedded PHP deployments, preserve volatile evidence on suspected compromised hosts, and apply PHP 8.5.12, 8.4.27, 8.3.36, or 8.2.35, or verified distribution backports; restart affected services and rebuild container images. Next, isolate systems that cannot be patched promptly and retain current WAF protections and tested request restrictions as temporary exposure reduction, not complete fixes. Finally, correlate suspicious requests with memory errors, worker instability, unusual response data, unexpected child processes, file changes, and outbound connections; investigate possible data disclosure as well as code execution, and rotate credentials exposed on compromised hosts.
Source: https://www.wiz.io/blog/php2shell-php-rce; https://www.crowdstrike.com/en-us/blog/crowdstrike-protects-customers-from-php-rce-zero-days/; https://www.imperva.com/blog/php2shell-vulnerability-updates/; https://www.armis.com/blog/php2shell-how-to-find-prioritize-and-patch-affected-systems/; https://blog.cloudflare.com/php-vulnerabilities-waf-protection/; https://github.com/php/php-src/security/advisories/GHSA-w8hw-w9fx-j79w; https://github.com/php/php-src/security/advisories/GHSA-6wmf-4rv5-p3rp
10/21/26
LiteLLM 1.107.0 Supply-Chain Credential Theft and Backdoor
HIGH
+ —
Intel Source:
StepSecurity
Intel Name:
LiteLLM 1.107.0 Supply-Chain Credential Theft and Backdoor
Date of Scan:
10/21/26
Impact:
HIGH
Summary:
Attackers distributed a credential stealer and persistent remote-code loader through LiteLLM 1.107.0 on PyPI and official container images, exposing developer systems, CI pipelines, and AI gateways. StepSecurity confirmed malicious additions in all seven compiled wheels across Windows, macOS, and Linux, with matching payloads in container builds. Python startup hooks execute the malware whenever an interpreter starts; importing LiteLLM is unnecessary. The stealer collects browser passwords, SSH and GPG keys, cloud credentials, running-process environments, and configuration data from more than 20 AI coding tools. On proxy deployments, it also targets provider credentials in the LiteLLM database. A background loader periodically consults a GitHub dead drop and executes attacker-supplied Python. Build provenance points to a lookalike repository, while the legitimate release tag remains clean. How the release workflow was redirected remains unresolved. Removal does not eliminate exposure from cached packages or previously pulled images held in internal registries. Prioritize identifying affected installations and image digests, stopping the loader, and replacing compromised containers with a verified clean release pinned by digest. Rotate every credential accessible to affected hosts, including proxy-database provider keys and cloud sessions. Block the reported attacker infrastructure and review historical egress for evidence of credential transmission or additional remote-code execution.
Source: https://www.stepsecurity.io/blog/litellm-1-107-0-compromised-pypi-backdoored-wheels-credential-stealer
10/21/26
k0t Network-Edge Exploitation and Corporate Access Campaign
HIGH
+ —
Intel Source:
Group-IB
Intel Name:
k0t Network-Edge Exploitation and Corporate Access Campaign
Date of Scan:
10/21/26
Impact:
HIGH
Summary:
Operators associated with k0t exploited exposed network appliances and weak credentials to enter corporate networks worldwide, establish persistent remote access, and steal data. Group-IB and NyxLab recovered more than 70,000 exposed credentials associated with network devices and directory information. Their investigation identified 84 compromised devices across 28 active organizational groups in an attacker-operated MeshCentral console; another 100 device groups had been removed. These figures describe recovered infrastructure and credentials, rather than a verified count of independently affected companies. Initial access combined exploitation of FortiWeb CVE-2025-64446, brute-force attacks against FortiGate authentication, and forged authentication-override cookies exploiting GlobalProtect CVE-2026-0257. After authentication, tooling could download appliance configurations and recover stored secrets, including credentials useful for access to internal directories. Operators used OpenConnect for VPN access, deployed MeshAgents and AdaptixC2 for persistence, and abused administrative tools to enumerate networks, extract credentials, and escalate privileges. In sufficiently compromised victims, the investigators observed Rclone transferring data to cloud storage. Some victims subsequently appeared on Qilin and The Gentlemen leak sites, but a direct relationship between the access operators and those ransomware groups remains unconfirmed. An AI penetration-testing tool was present in the recovered tooling without evidence of deployment against victims. First, identify exposed FortiGate, FortiWeb, and GlobalProtect systems, remediate the reported vulnerabilities, and restrict management access. Next, review appliance administration, VPN authentication, and unexpected remote-management installations for unauthorized access. Assess and rotate credentials stored on or reachable through compromised appliances, extending the investigation into internal systems where those credentials could have been used.
Source: https://www.group-ib.com/blog/stray-cat-network-edge/
10/21/26
Amazon Business Search-Ad ClickFix Malware Campaign
MEDIUM-HIGH
+ —
Intel Source:
Huntress
Intel Name:
Amazon Business Search-Ad ClickFix Malware Campaign
Date of Scan:
10/21/26
Impact:
MEDIUM-HIGH
Summary:
An Amazon-themed ClickFix campaign infected endpoints belonging to at least three Huntress customers after users followed a sponsored Google result for an Amazon Business search. The result redirected visitors to a spoofed Amazon login page hosted through Google Sites. A fake CAPTCHA copied a command to the clipboard and instructed users to paste it into the Windows Run dialog. The command invoked the Windows Finger utility to retrieve script content, beginning an obfuscated, multistage infection chain. The attack downloaded a legitimate IronPython distribution, used encoded Python and encrypted payloads to load code into memory, and deployed a Node.js component that retrieved additional malware. Huntress observed a payload assessed as Lumma Stealer alongside Doina and an unauthorized, older NetSupport Client installation. The stealer profiled the host and received instructions concerning browsers, password managers, email clients, and cryptocurrency applications. NetSupport supplied a separate remote-access channel. Payload archives changed hashes through added padding, and their contents changed once or twice daily, limiting the usefulness of static archive hashes alone. The page also offered alternate macOS instructions, but that delivery location was offline during the investigation; the detailed infection evidence concerns Windows. Google removed the malicious advertisement, while Huntress reported that attacker infrastructure remained functional at research time. Prioritize teaching users to reject CAPTCHA requests that require operating-system commands, and restrict the Run dialog where operationally appropriate. Investigate endpoint and network telemetry for the reported delivery chain and associated infrastructure, using browser history and advertising-redirect context to reconstruct initial access. Report the malicious advertising evidence to Google to interrupt further distribution.
Source: https://www.huntress.com/blog/amazon-clickfix-finger
10/21/26
BUGSTER AI-Assisted Credential Theft and University Data Exfiltration
MEDIUM-HIGH
+ —
Intel Source:
Palo Alto Networks
Intel Name:
BUGSTER AI-Assisted Credential Theft and University Data Exfiltration
Date of Scan:
10/21/26
Impact:
MEDIUM-HIGH
Summary:
A financially motivated operator using the BUGSTER handle combined AI-generated reconnaissance and attack scripts with weak credentials to attack organizations in India, including two universities. Unit 42 confirmed that the actor exfiltrated more than 5,200 records containing personal information from one university's MongoDB database. At another university, the operator enumerated exposed services and WordPress authors, generated candidate usernames, and attempted password guessing. The tooling flagged a candidate login, but that result does not establish the same level of compromise as the confirmed database theft. The actor also enumerated an Indian e-commerce site and collected exposed administrative contact information and a storefront API token. The operational environment centered on Hermes Agent with DeepSeek V4, controlled through Telegram, and a Cursor development environment configured to bypass permission and sandbox controls. Specialized bot profiles supplied advice and retained operational knowledge, reducing the expertise required to develop tools and pursue different objectives. The recovered server also supported a command-and-control listener and reverse-shell payloads. Other material concerned VoIP fraud, Wi-Fi password cracking, and cryptocurrency reconnaissance; the report does not establish completed VoIP fraud or follow-on attacks against the identified cryptocurrency holders. The evidence therefore supports confirmed data theft and reconnaissance without treating every recovered capability as a successful operation. Prioritize investigating connections to the reported infrastructure and correlating suspicious endpoint activity with command-and-control behavior. Strengthen privileged access through phishing-resistant MFA and least privilege, and secure exposed application and database credentials. Assess enterprise AI agents, their permissions, and secret access, retaining controls that prevent unauthorized execution and credential disclosure.
Source: https://unit42.paloaltonetworks.com/ai-assisted-cybercrime-analysis/
10/21/26
Fake Amazon Music Sponsorship Phishing Against YouTube Creators
MEDIUM
+ —
Intel Source:
Malwarebytes
Intel Name:
Fake Amazon Music Sponsorship Phishing Against YouTube Creators
Date of Scan:
10/21/26
Impact:
MEDIUM
Summary:
A phishing operation used fabricated Amazon Music sponsorship offers to solicit Google credentials from YouTube creators, placing channel ownership and associated account access at risk. Malwarebytes found a wave of sites presenting an invitation-only creator portal that borrowed credibility from the legitimate Amazon and YouTube shopping partnership. Visitors entered a channel link and received a personalized dashboard containing real subscriber statistics, uploads, playlists, and thumbnails. The site collected this public information through YouTube's API and rotated among 20 embedded API keys when limits or errors occurred. An earnings calculator and customizable campaign packages encouraged visitors to continue to a purported ownership-verification step. Although genuine Google sign-in code was present, its configuration was inactive. The visible sign-in button instead opened an almost transparent frame containing a credential form served from separate attacker infrastructure. The form transmitted entered credentials to its own server and supported MFA challenge steps. A subsequent personalized offer and request-confirmation screen could reassure visitors after submission, delaying recognition of the phishing attempt. Malwarebytes identified signs of a reusable kit but could not confirm how it was shared. The reporting establishes malicious credential-collection infrastructure and distribution, without providing a confirmed victim count or demonstrating completed channel takeovers. Prioritize independent verification of sponsorship outreach and require users to check the browser's actual sign-in location before supplying passwords or verification codes. For suspected submissions, change the Google password, terminate unfamiliar sessions, and review recovery settings and third-party access. Check YouTube and Brand Account permissions for unauthorized administrators, and use the platform's account-recovery process if access has been lost.
Source: https://www.malwarebytes.com/blog/threat-intel/2026/10/fake-amazon-music-x-youtube-deal-lures-creators-into-a-google-login-trap
10/21/26
Staples Order-Notification Abuse for Callback Fraud
MEDIUM
+ —
Intel Source:
Cisco Talos
Intel Name:
Staples Order-Notification Abuse for Callback Fraud
Date of Scan:
10/21/26
Impact:
MEDIUM
Summary:
Attackers abused Staples' transactional email workflow to deliver fraudulent purchase notifications containing an attacker-controlled callback number. Cisco Talos analyzed this campaign alongside a separate Dell notification campaign; together, the two accounted for approximately 1,400 observed messages, with no campaign-specific breakdown provided. The Staples messages resembled ordinary order confirmations for an Apple iPad, complete with delivery information, pricing, billing details, and standard retailer footer content. Attackers inserted purported helpline instructions into fields normally used for shipping or billing information. The resulting message encouraged recipients to resolve an unexpected expensive purchase by telephone, moving the interaction outside the email channel. Talos reported aligned DKIM and passing DMARC results for the visible retailer identity, alongside forwarding through Microsoft 365 and other enterprise mail infrastructure. Those authentication results supported the credibility of the delivery wrapper while leaving the attacker-influenced transaction content unvalidated. Talos documented deceptive notifications directing recipients toward attacker-controlled contact information. It did not report completed fraudulent payments or remote access following a callback, or establish a compromise of the retailer's email systems. Prioritize correlating unexpected order messages with actual account activity, payment state, and independently obtained support contacts before responding. Platform operators should validate user-supplied transaction fields before storing and rendering them, flagging misplaced support instructions, phone numbers, and urgency language. Mail-security teams should inspect the fully rendered message and evaluate whether its content fits the transaction workflow, rather than allowing authentication results or a trusted template to determine the disposition alone.
Source: https://blog.talosintelligence.com/parasitizing-trust-how-legitimate-infrastructure-becomes-a-delivery-pipeline-for-attackers/
10/21/26
Dell Order-Notification Abuse for Callback Fraud
MEDIUM
+ —
Intel Source:
Cisco Talos
Intel Name:
Dell Order-Notification Abuse for Callback Fraud
Date of Scan:
10/21/26
Impact:
MEDIUM
Summary:
A callback-fraud campaign placed attacker-controlled telephone instructions inside Dell order-received notifications, using the retailer's familiar purchase workflow to make deceptive messages appear credible. Cisco Talos observed messages presenting an order for Microsoft Project Standard 2024 valued at $736.09. Shipping and billing sections contained order-confirmation language, a purported helpline, and an instruction to call a sales team. This combination framed an unexpected purchase as an urgent problem and positioned the attacker's phone number as the resolution channel. The campaign used the same broad technique as a separately described Staples campaign, but the report does not establish common operators. Talos counted approximately 1,400 messages across both campaigns without assigning a separate Dell total. Authentication analysis found passing DKIM and DMARC for Dell-related infrastructure, with forwarding and sender-rewriting artifacts as messages traversed Microsoft and enterprise mail systems. The malicious element resided in the rendered transaction fields, so authentic delivery did not validate the order or the embedded contact information. The report documents deceptive message distribution, rather than confirmed financial losses, credential theft, or compromise following a phone call. Enterprise mail defenses can encounter attacker-supplied transaction content within messages from expected senders using legitimate templates. First, verify unexpected purchase notifications against account and payment records, using independently established retailer contact channels. For platform owners, reject or flag inappropriate telephone instructions and urgent support text in transaction fields. For security operations, inspect complete rendered messages and correlate their content with the expected business workflow, escalating mismatched transaction details even when sender authentication passes.
Source: https://blog.talosintelligence.com/parasitizing-trust-how-legitimate-infrastructure-becomes-a-delivery-pipeline-for-attackers/
10/20/26
SafePay Ransomware and Data Extortion Across Critical Infrastructure
HIGH
+ —
Intel Source:
Cybersecurity and Infrastructure Security Agency
Intel Name:
SafePay Ransomware and Data Extortion Across Critical Infrastructure
Date of Scan:
10/20/26
Impact:
HIGH
Summary:
CISA reports that SafePay ransomware continues to compromise businesses and critical infrastructure, combining data theft, encryption, and recovery sabotage to increase extortion pressure. A joint advisory describes investigations through August 2026 and reports at least 600 separate victims across North America, South America, Europe, and Australia. Operators gain access through stolen or guessed credentials, phishing, and exploitation of vulnerable edge and backup systems. They establish administrative access, repurpose legitimate management utilities for lateral movement, and enumerate network shares to locate valuable information. Before encryption, they compress and transfer data through legitimate file transfer and cloud services, disable security software, and interfere with backups. Reported recovery disruption includes stopping shadow copy services and resetting network storage that holds backup data, while some victims also receive telephone pressure after data theft and encryption. Defenders should prioritize patching exposed gateways and backup products, reduce unnecessary remote access, and require multifactor authentication with restricted administrative privileges. Investigate unexpected accounts, remote management activity, security service changes, and unusual outbound transfers, then isolate confirmed compromised hosts and scope affected credentials. Maintain segmented, immutable or offline backups and test restoration procedures so recovery remains possible after production systems and connected storage are affected.
Source: https://www.cisa.gov/news-events/cybersecurity-advisories/aa26-293a
10/20/26
OVHcloud CLI Release Asset Compromise and Developer Secret Theft
HIGH
+ —
Intel Source:
OVHcloud
Intel Name:
OVHcloud CLI Release Asset Compromise and Developer Secret Theft
Date of Scan:
10/20/26
Impact:
HIGH
Summary:
OVHcloud confirmed that an attacker replaced command line client release assets with credential-stealing binaries, creating a supply chain exposure for developer workstations and cloud administration systems. A compromised developer account replaced assets for versions 0.11.0 through 0.15.0 during a brief window on October 19, 2026, and approximately 250 downloads involved affected Linux, macOS, and Windows builds. The malicious client collects cloud credentials, deployment configurations, SSH keys, and data from development tools and AI coding assistants before preparing the collection for exfiltration. OVHcloud states that source code, the build chain, other repositories, and internal production development systems were unaffected. It removed the malicious assets, suspended the compromised account, enabled immutable releases, and published version 0.16.0 with a signed tag. Organizations should identify systems that downloaded and executed the client during the affected window, stop using suspect copies, and replace them with the clean release. Revoke or rotate all secrets accessible from affected systems and investigate associated cloud and repository access. Require release integrity checks, restrict publishing privileges, and enforce immutable distribution assets to reduce the risk of post-publication replacement.
Source: https://github.com/ovh/ovhcloud-cli/security/advisories/GHSA-862f-55f5-vmvw
10/20/26
GlobalProtect Authentication Bypass Enables Ransomware Intrusions
HIGH
+ —
Intel Source:
Palo Alto Networks
Intel Name:
GlobalProtect Authentication Bypass Enables Ransomware Intrusions
Date of Scan:
10/20/26
Impact:
HIGH
Summary:
Palo Alto Networks reports that ransomware operators are exploiting a PAN-OS GlobalProtect authentication bypass to establish unauthorized VPN access and move into enterprise infrastructure. An October 19 research update identifies Qilin and INC operators among those using the vulnerability across business and public sector environments. Attackers bypass multifactor authentication, conduct internal reconnaissance with remote management tools, and move laterally through remote desktop and file sharing services. Subsequent activity includes credential theft, exfiltration, and ransomware deployment. In one investigated intrusion, attackers progressed from a compromised gateway to virtualization management, disabled hypervisor protections, and extracted credentials directly from a domain controller virtual disk. This approach can evade security tools operating inside the guest system and expose credentials with wider domain privileges, and researchers also observed exploitation tooling targeting backups. Defenders should prioritize upgrading affected gateways to a fixed release or applying vendor workarounds and mitigations. Examine VPN authentication and connection logs for successful unauthorized sessions, activate incident response when compromise is identified, and extend investigation into remote administration and lateral movement. Review virtualization and backup management access, restore disabled protections, and rotate credentials exposed through compromised systems or virtual disks.
Source: https://unit42.paloaltonetworks.com/active-exploitation-of-pan-os-cve-2026-0257/
10/20/26
INC and Gentlemen Affiliates Operate AI-Assisted Intrusion Infrastructure
HIGH
+ —
Intel Source:
SOCRadar
Intel Name:
INC and Gentlemen Affiliates Operate AI-Assisted Intrusion Infrastructure
Date of Scan:
10/20/26
Impact:
HIGH
Summary:
SOCRadar reports that exposed infrastructure reveals an operation supporting INC and The Gentlemen ransomware affiliates, with AI assisting reconnaissance, exploitation, credential handling, and extortion planning. Researchers analyzed seven servers containing session histories, knowledge bases, tooling, and victim records. Evidence includes ransomware preparations for six victims and active encryption, although deployment of the full wave remains unconfirmed. Operator records describe exploitation of FortiClient EMS, SimpleHelp, cPanel, and other remote access systems, but the reported exploitation totals were not independently verified. The research identifies 62 organizations at differing stages of targeting or compromise, including 36 in the United States. Reused credentials enabled access across management platforms, while reverse tunnels supported collection of sensitive files. SOCRadar assesses the evidence as supporting shared infrastructure more strongly than a formal alliance between the brands. Defenders should prioritize patching exposed management and remote access products, restrict their interfaces, and audit new privileged accounts and configuration changes. Strengthen managed service access with multifactor authentication and separate administrative credentials, investigate unauthorized remote tools and bulk outbound transfers, and protect backups with isolated or immutable storage. Enable endpoint tamper protection and prepare a coordinated response to encryption and data extortion.
Source: https://socradar.io/blog/inc-gentlemen-ransomware-ai-infrastructure/
10/20/26
Forged Commits Add Malicious Install Hooks to Developer Projects
MEDIUM-HIGH
+ —
Intel Source:
SafeDep
Intel Name:
Forged Commits Add Malicious Install Hooks to Developer Projects
Date of Scan:
10/20/26
Impact:
MEDIUM-HIGH
Summary:
SafeDep reports that forged repository commits introduced malicious installation hooks into a Hugo theme and a JavaScript application scaffolding project, exposing developers and build runners to attacker-controlled code execution. Researchers found pushes copying legitimate author identities, dates, and messages to disguise malicious changes as established project history. One repository had a signed merge replaced with an unsigned commit, while the other received commits carrying copied metadata. The poisoned branches execute a downloader when dependencies are installed with lifecycle scripts enabled, selecting a second stage for Windows, macOS, or Linux. Registry releases remained clean, but subsequent forks or copies could inherit the payload. The second stage was not retrieved, so its ultimate capabilities remain unknown, and SafeDep separates this activity from earlier campaigns using similar commit forgery. Defenders should identify repository-based installations and inherited forks, verify branch history and signatures, and prevent unreviewed installation scripts from running. Treat machines and build runners that executed the hook as compromised, investigate their process and outbound activity, and assess the full accessible environment. Rotate repository, publishing, cloud, and other reachable credentials before returning affected development systems to trusted use.
Source: https://safedep.io/hugo-paper-mevn-cli-forged-commit-preinstall-dropper/
10/20/26
Swallowtail Espionage Intrusions Against Defense and Law Enforcement
MEDIUM-HIGH
+ —
Intel Source:
Symantec
Intel Name:
Swallowtail Espionage Intrusions Against Defense and Law Enforcement
Date of Scan:
10/20/26
Impact:
MEDIUM-HIGH
Summary:
Symantec reports that Swallowtail espionage intrusions targeted defense and law enforcement organizations, with sustained access and credential theft in environments holding sensitive national security information. Researchers identified activity against a naval shipbuilder, a military aircraft engine manufacturer, and a law enforcement agency in two unnamed NATO member states. One intrusion lasted at least two months, with operators repeatedly returning to a compromised host. An email attachment delivered Hookedge through an obfuscated Word macro, and a hidden browser retrieved tasks and returned command output. Later activity moved to dedicated attacker infrastructure. Operators established a persistent SSH tunnel into the internal network and captured authentication material through a local listener that induced an NTLM handshake. They also deployed an undocumented reflective loader, but its encrypted payload was not recovered and its ultimate function remains unknown. Defenders should prioritize controls on macro-enabled attachments and investigate unusual browser launches from document or script processes. Extend incident review to SSH tunnels, authentication capture, scheduled persistence, and repeated interactive access rather than stopping at the initial backdoor. Contain affected hosts, assess credential exposure, and remove persistent access before restoring systems to trusted operation.
Source: https://www.security.com/threat-intelligence/swallowtail-russia-nato-targets
10/20/26
Cairn-Assisted Intrusions Target Commerce and Payment Systems
MEDIUM-HIGH
+ —
Intel Source:
NorForge
Intel Name:
Cairn-Assisted Intrusions Target Commerce and Payment Systems
Date of Scan:
10/20/26
Impact:
MEDIUM-HIGH
Summary:
NorForge reports that criminal operators use the Cairn offensive agent framework to pursue persistent access and financial theft against commerce and payment systems, with researchers validating intrusions. Researchers examined exposed control panels and distinguished multiple clusters, including two assessed as malicious with high confidence. One cluster managed more than 1,100 projects targeting shopping sites in the United States, Canada, Japan, and Australia, while another focused on stealing payment data and cryptocurrency. Tasks combined reconnaissance with authentication, unsafe upload, injection, and deserialization attacks, often seeking persistent access. Researchers confirmed an uploader on a compromised vendor site and described additional validated compromise cases. Exposed panels also contained credentials, including secrets verified as valid, creating further risk for targets. Success estimates and some operator claims remain uncertain. Defenders should prioritize remediation of exposed commerce applications, restrict administrative and data services, and address unsafe upload and deserialization behavior. Investigate unexpected server files, account changes, and application or database access, then isolate confirmed compromised systems and rotate exposed secrets. Secure legitimate offensive testing platforms with authenticated, restricted access and prevent assessment logs from exposing client credentials or sensitive findings.
Source: https://norforge.com/blog/inside_cairn
10/19/26
ConnectWise Vendor Credential Abuse and Microsoft 365 Persistence
HIGH
+ —
Intel Source:
Huntress
Intel Name:
ConnectWise Vendor Credential Abuse and Microsoft 365 Persistence
Date of Scan:
10/19/26
Impact:
HIGH
Summary:
Huntress reports that attackers abused compromised ConnectWise application credentials to access Microsoft 365 tenants, creating a supply-chain risk for managed service providers and their customers. The investigation confirmed unauthorized application sign-ins and identified two tenants where the attacker quickly obtained Global Administrator privileges and planted credentials on other applications, which could preserve access after the original vendor applications were disabled. In its October 19 update, Huntress found that the credential planting targeted locally registered applications without a consistent preference for their privileges. Huntress found no evidence of mailbox or file manipulation, or user or policy changes, in those two tenants, while cautioning that some directory reads leave no audit record. Its last confirmed successful sign-in using the vendor credential was on October 15, followed by failed attempts within the telemetry available to Huntress. Defenders should disable affected ConnectWise applications without deleting them and confirm vendor credential resets before restoring access. Audit credentials and privileged roles across all tenant applications, remove unauthorized additions, and investigate service principal sign-ins. Review workload identity controls and application permission boundaries, while accounting for interrupted backups during containment.
Source: https://www.huntress.com/blog/vendor-credential-enables-unauthorized-access
10/19/26
NetScaler Exploit Chain and Persistent Appliance Intrusions
HIGH
+ —
Intel Source:
Rapid7
Intel Name:
NetScaler Exploit Chain and Persistent Appliance Intrusions
Date of Scan:
10/19/26
Impact:
HIGH
Summary:
Rapid7 reports that attackers chained two NetScaler vulnerabilities to gain root execution and turn an exposed enterprise gateway into a persistent platform for reconnaissance, tunneling, and configuration theft. The investigation involved a compromised appliance running an affected 14.1 build and correlated authentication log poisoning, service crashes, reboot activity, and subsequent root payload execution. The intrusion chained CVE-2026-88771 with CVE-2026-88779, using a crash and reboot to trigger retained attacker-controlled command text. Operators deployed web shells, redundant command channels, scheduled execution, and boot persistence, while attempting to synchronize access across a high-availability pair. A dedicated reconnaissance tool examined internal identity services, management systems, and virtualization infrastructure, and Rapid7 confirmed configuration exfiltration that could expose credentials, private keys, and other sensitive trust information. Organizations should apply the relevant fixed builds and investigate appliances that were exposed while vulnerable. Preserve volatile evidence before disruptive changes and examine both high-availability nodes for unauthorized persistence, privilege changes, and internal scanning. Rotate exposed credentials and cryptographic material, then verify that remediation removed the attacker foothold across the appliance pair and connected systems.
Source: https://www.rapid7.com/blog/post/ra-netscaler-intrusions-chaining-cve-2026-88771-and-cve-2026-88779/
10/19/26
Slaiclind ClickFix and EtherHiding Malware Campaign
MEDIUM-HIGH
+ —
Intel Source:
Stormshield
Intel Name:
Slaiclind ClickFix and EtherHiding Malware Campaign
Date of Scan:
10/19/26
Impact:
MEDIUM-HIGH
Summary:
Stormshield reports that the Slaiclind campaign combines compromised websites, blockchain-based payload retrieval, and fake verification prompts to induce malware execution on Windows endpoints. The recurring script, observed from late September, retrieves attacker-controlled delivery locations from Polygon smart contracts. The injected website code avoids displaying its activity to certain WordPress administrators, while the subsequent lure targets Windows users and filters search-engine bots. A fake CAPTCHA places a command on the clipboard and directs the user toward shell execution, after which the chain installs a legitimate runtime, downloads additional code, and attempts process ghosting and injection into a Windows shell process. Stormshield analyzed two payloads, including an executable and encoded shellcode, but did not establish enterprise victim counts or a specific geographic target. Defenders should prioritize controls and user reporting that interrupt unexpected website instructions to paste commands into a terminal. Restrict unapproved runtime installation and script execution, and examine unusual browser-to-shell execution chains. Investigate process injection and suspicious payload activity promptly, isolating affected endpoints and restoring confirmed compromised systems from trusted sources.
Source: https://www.stormshield.com/news/cti-payload-slaiclind-campaign/
10/16/26
GhostAction Developer Credential Theft and Persistence
HIGH
+ —
Intel Source:
OpenSourceMalware
Intel Name:
GhostAction Developer Credential Theft and Persistence
Date of Scan:
10/16/26
Impact:
HIGH
Summary:
OpenSourceMalware reports that GhostAction has expanded its repository-secret theft campaign to compromise developer workstations, creating persistent access that can survive removal of malicious workflows. The researchers identified 94 compromised repositories belonging to 66 owners and analyzed 79 accessible repositories, all of which carried the second stage. The attacker added a hidden editor task and a fake font containing credential-stealing code. Execution requires either a trusted Visual Studio Code workspace with automatic tasks permitted or direct execution of the disguised payload. The payload collects cloud credentials, package-registry tokens, SSH keys, and AI-tool credentials, then installs recurring command retrieval on macOS and Linux. Windows systems remain exposed to credential theft even though that persistence mechanism does not install there. Two repository owners removed earlier malicious workflows but were compromised again, showing that workflow removal alone did not end attacker access. OpenSourceMalware distinguishes this activity from PolinRider because the payload and infrastructure differ. Defenders should contain potentially affected workstations and revoke compromised repository sessions, tokens, application grants, and keys from a clean device. Audit every writable repository, branch, and fork for unauthorized workflows and editor automation, inspect developer systems for persistence, rotate exposed credentials after containment, and require review and approval for workflow changes and execution.
Source: https://opensourcemalware.com/blog/ghostaction-copycats-polinrider-by-hiding-a-stealer-in-a-fake-font-file
10/16/26
GitLab CI/CD Regex Integer Overflow Code Execution
HIGH
+ —
Intel Source:
Rapid7
Intel Name:
GitLab CI/CD Regex Integer Overflow Code Execution
Date of Scan:
10/16/26
Impact:
HIGH
Summary:
Rapid7 demonstrated command execution through a GitLab CI/CD regular-expression integer overflow, creating a serious risk to self-managed development infrastructure. The analysis covers CVE-2026-97229 in GitLab Community and Enterprise editions across affected 19.2, 19.3, and 19.4 releases. Researchers reproduced execution as the GitLab operating-system account on the official Linux amd64 package with address randomization enabled. Their demonstrated route requires the attacker to place malicious pipeline configuration and a specially constructed repository path in a public project before an unauthenticated validation request triggers execution. Rapid7 notes that this prerequisite differs from the vendor severity description and that the underlying private report may describe another route. The vendor release notes did not report exploitation in the wild at disclosure. Successful execution can crash an application worker, while constrained memory can interrupt the attempt. Defenders should upgrade affected self-managed installations to 19.2.8, 19.3.5, 19.4.2, or an appropriate later fixed release. While upgrades are underway, restrict unauthenticated access to attacker-writable public projects or the vulnerable validation function. Investigations should correlate application and proxy errors with worker restarts, abnormal memory growth, unexpected child processes, and suspicious repository changes when assessing possible compromise.
Source: https://www.rapid7.com/blog/post/ra-gitlab-ci-cd-regex-integer-overflow-technical-analysis-cve-2026-97229/
10/16/26
RuRAT Persistent Windows Remote Access
MEDIUM
+ —
Intel Source:
CYFIRMA
Intel Name:
RuRAT Persistent Windows Remote Access
Date of Scan:
10/16/26
Impact:
MEDIUM
Summary:
CYFIRMA describes a malicious Windows sample tracked as RuRAT that combines persistent remote control with credential collection, allowing an operator to retain access to compromised enterprise endpoints. Its weekly analysis identifies Windows as the affected technology and describes worldwide targeting, but provides no quantified victim count or campaign scale. The analyzed sample uses scheduled tasks, logon startup configuration, and Windows services to survive routine system restarts and user sessions. It also exhibits memory-based execution and code injection that can obscure payload inspection and place activity inside trusted processes. The report describes a continuing communication channel through which a remote operator can issue commands and receive collected information. These capabilities support repeated access rather than a single theft event, although the report does not establish a particular initial-access route for this sample. Defenders should isolate confirmed infected endpoints and investigate unauthorized remote administration, new services, scheduled tasks, and startup changes before restoring access. Reset affected account credentials, revoke active sessions, and verify multifactor authentication after containment. Review endpoint process activity and security-setting changes, maintain consistent protection coverage, and restore systems only after persistence and unauthorized access have been removed.
Source: https://www.cyfirma.com/news/weekly-intelligence-report-16-oct-2026/
10/15/26
skynox7 Proxyjacking Campaign
HIGH
+ —
Intel Source:
GreyNoise
Intel Name:
skynox7 Proxyjacking Campaign
Date of Scan:
10/15/26
Impact:
HIGH
Summary:
GreyNoise reports that the skynox7 actor compromised at least 143 internet-facing edge devices across 34 countries and installed residential proxy software, turning victim systems into proxy infrastructure. The campaign exploited known weaknesses in network storage, video recording, and enterprise application platforms, with backup remote access established on some devices. This activity can conceal later intrusion traffic behind residential connections and make normal location-based authentication checks less reliable. Compromised devices may also provide attackers with durable network footholds and bandwidth that can support fraud or other malicious activity. GreyNoise also describes a related enterprise intrusion investigation in which multiple users authenticated from the same residential address, a pattern that warranted urgent investigation. Patch affected edge devices and remove public exposure where it is not required. Review identity and remote-access logs for shared residential-origin activity, abnormal device or session characteristics, and unexpected administrative access. Audit device configurations for unauthorized remote access, preserve relevant evidence, and rebuild compromised systems from trusted sources.
Source: https://www.greynoise.io/blog/proxyjacking-edge-devices-residential-proxies
10/15/26
YuanKong Android Remote Access Trojan
MEDIUM-HIGH
+ —
Intel Source:
Star Labs
Intel Name:
YuanKong Android Remote Access Trojan
Date of Scan:
10/15/26
Impact:
MEDIUM-HIGH
Summary:
Star Labs documents YuanKong, an Android remote access trojan delivered through droppers that unpack a shared, multi-tenant remote-control payload. The research identified 110 droppers, 57 distinct child payloads, and 42 operator accounts, with each build carrying operator-linked configuration and a shared family signing identity. YuanKong can capture screens, inspect accessibility content including password fields, collect notifications and clipboard data, control the device remotely, and target payment PIN entry. The report also describes abuse of Android wireless debugging to gain a local shell and grant additional permissions, increasing persistence and device-control risk. Restrict application installation to managed sources and investigate Android devices that request unusual accessibility, screen-capture, or wireless-debugging access. Hunt mobile telemetry for the documented family signals and review endpoint records for unauthorized permission changes, remote-control behavior, or persistent background services. Revoke affected credentials, reset payment access where appropriate, and rebuild or re-enroll compromised devices through trusted management workflows.
Source: https://starlabs.sg/blog/2026/10-yuankong-rat-the-android-remote-control-service-that-stamps-its-operators-names-into-every-build/
10/15/26
Midnight Blizzard-linked ChocoShell and CornFlake Activity
HIGH
+ —
Intel Source:
ReversingLabs
Intel Name:
Midnight Blizzard-linked ChocoShell and CornFlake Activity
Date of Scan:
10/15/26
Impact:
HIGH
Summary:
ReversingLabs expanded two previously published samples into a Midnight Blizzard-linked campaign involving phishing delivery, PowerShell staging, credential theft, and a Go backdoor. The activity targeted more than 20 government, diplomatic, think-tank, and defense-industrial organizations, mostly in Ukraine and Europe. ReversingLabs identified additional stagers, lure files, related infrastructure, and detection content, and reported that the staging system was still serving a malicious file on October 5. The stager can collect browser data and execute hidden PowerShell, while the backdoor establishes persistence and accesses browser profiles and collaboration application data. Restrict device-code authentication where business use permits and investigate Entra ID sign-ins associated with abnormal device-code flows or suspicious consent activity. Deploy and validate the publisher-provided hunting rules, then review PowerShell, process-creation, persistence, browser-data access, and network telemetry for matching behavior. Block confirmed malicious infrastructure through approved controls, rotate exposed cloud credentials, and remove affected endpoints from service for trusted rebuild and forensic review.
Source: https://www.reversinglabs.com/blog/how-rl-mapped-a-midnight-blizzard-linked-campaign-from-two-hashes
10/14/26
YAM Store and Iron Man phishing ecosystem
HIGH
+ —
Intel Source:
Microsoft
Intel Name:
YAM Store and Iron Man phishing ecosystem
Date of Scan:
10/14/26
Impact:
HIGH
Summary:
Microsoft uncovered an interconnected phishing ecosystem that compromises cloud environments and supplies criminal operators with infrastructure for large-scale credential theft and financial fraud. YAM Store brokers compromised or fraudulently obtained Microsoft and Google environments, while related phishing kits target healthcare, financial services, wholesale distribution, and information technology organizations. Since January 2026, the kit operators have facilitated more than one billion phishing emails and text messages targeting over 75,000 organizations worldwide. Stolen application credentials can provide durable directory access, including password resets and license assignment, while compromised storage is repurposed as phishing infrastructure. Iron Man supports live victim interaction and the collection of credentials, payment information, and verification codes, and downstream operators also use device-code phishing to obtain account access. Microsoft coordinated a disruption that affected the ecosystem, but the report does not establish that all related operations have ended. Defenders should prioritize phishing-resistant authentication, conditional access, and restricted administrator privileges while strengthening email and web protections. Require independent verification of urgent authentication or payment requests, audit privileged applications and cloud permissions, and investigate unusual account activity. Revoke compromised access and secure affected identities when exposure is confirmed.
Source: https://www.microsoft.com/en-us/security/blog/2026/10/14/the-hidden-network-inside-the-phishing-infrastructure-ecosystem-powering-global-cybercrime/
10/14/26
Trojanized AI installers steal browser sessions
HIGH
+ —
Intel Source:
ANYRUN-FZCO
Intel Name:
Trojanized AI installers steal browser sessions
Date of Scan:
10/14/26
Impact:
HIGH
Summary:
ANYRUN-FZCO documented trojanized AI application installers that steal browser sessions while still installing genuine software, creating a route into enterprise email, customer-management systems, and cloud consoles. In the examined fake Claude download sequence, a malicious setup program launches the authentic installer as a decoy while a hidden stager creates recurring execution of Microsoft Edge with remote debugging enabled. The browser then exports cookies and extension data to attacker-controlled infrastructure, exposing active sessions without requiring a fresh password or multifactor prompt. Shared registration behavior also appeared behind a different AI-branded application, indicating reuse of the installation kit. Successful installation of the legitimate application therefore does not establish that the downloaded wrapper was safe. The report provides no verified victim count or targeting geography for this specific installer activity. Defenders should prioritize managed AI software distribution and verify the provenance of the initial download, including wrappers obtained from public release platforms. Monitor installation-to-browser process chains, unusual background browser settings, and newly created recurring tasks. Contain suspicious endpoints and revoke exposed sessions and credentials before restoring business access.
Source: https://any.run/cybersecurity-blog/fake-ai-tools-us-companies/
10/14/26
Qilin Rust encryptor threatens backups and reliable recovery
HIGH
+ —
Intel Source:
Dark Atlas
Intel Name:
Qilin Rust encryptor threatens backups and reliable recovery
Date of Scan:
10/14/26
Impact:
HIGH
Summary:
Dark Atlas analyzed a Qilin Windows encryptor whose backup interference and partial file encryption can complicate enterprise containment and recovery. The Rust implementation supports configurable process and service interference, local and network-share traversal, and an attempt to delete shadow copies before encryption. Access to remote files depends on existing permissions and provider behavior, so the analysis does not establish autonomous propagation. File size affects which portions are encrypted, and implementation details can result in more encrypted data than the nominal range suggests. A failed rename can leave altered files under their original names, while interrupted operations may leave incomplete recovery metadata, meaning successful-operation counters do not reliably measure actual damage. The report separates this sample's behavior from broader ransomware-family activity and does not establish the affiliate's initial access or confirm data theft solely from ransom messaging. Defenders should prioritize isolation of affected systems and correlate rapid file changes with backup, process, and service interference. Preserve altered files and recovery metadata, including files whose names remain unchanged, before attempting restoration. Investigate compromised credentials and remote access separately, and restore from verified clean backups only after the intrusion path is addressed.
Source: https://darkatlas.io/blog/blog-qilin-ransomware-analysis
10/14/26
Zimbra account flaws expose mailboxes and persistent forwarding
MEDIUM-HIGH
+ —
Intel Source:
Rapid7
Intel Name:
Zimbra account flaws expose mailboxes and persistent forwarding
Date of Scan:
10/14/26
Impact:
MEDIUM-HIGH
Summary:
Rapid7 disclosed five Zimbra account vulnerabilities that weaken authentication controls, enable sender impersonation, expose other users' mailboxes, and permit persistent mail forwarding. The research covers an authentication rate-limit bypass, an external-directory injection flaw, unauthorized local sender selection, cross-user Exchange Web Services access, and a server-side filtering policy bypass. Several impacts require an ordinary compromised account, and mailbox access also depends on the relevant compatibility service being enabled. The mailbox flaw exposes messages, attachments, contacts, and calendar data, while the filtering flaw can send future message text and headers outside the organization. Saved forwarding rules may continue operating after a password change and can remain absent from the Modern client's filter view. The article reports researcher testing and does not establish active malicious exploitation. Three findings have published fixes, while two authentication findings lacked a publicly listed fixed release at publication. Administrators should apply supported security updates and track the remaining fixes, reduce unnecessary service exposure, and investigate abnormal authentication and cross-user mailbox activity. Inspect stored incoming and outgoing filters directly, remove unauthorized forwarding, and assess affected accounts and server-side state beyond the initial password reset.
Source: https://www.rapid7.com/blog/post/ve-breaking-zimbra-identity-boundaries-part-1/
10/14/26
Lorem Ipsum and Axolotl loaders use ClickFix delivery
MEDIUM-HIGH
+ —
Intel Source:
Forescout
Intel Name:
Lorem Ipsum and Axolotl loaders use ClickFix delivery
Date of Scan:
10/14/26
Impact:
MEDIUM-HIGH
Summary:
Forescout examined two malware-loader families whose deceptive installation chains and persistence create a route to deeper enterprise compromise. A late-September incident used ClickFix instructions to deliver Lorem Ipsum and a Python tunneling client, and investigators stopped the intrusion before ransomware deployment. The research compares incident samples with multiple Axolotl variants and documents additional delivery through trojanized Windows installers and layered scripting. Lorem Ipsum uses legitimate programs to load malicious libraries, injects into other processes, and establishes recurring execution through startup settings and scheduled tasks. Axolotl variants use different host programs, while both families share elements of their command-and-control registration structure. Their infrastructure resolution, persistence, and evasion methods also differ, limiting reliance on a single detection approach across every variant. Historical tradecraft links do not establish who operated the investigated incident. Defenders should prioritize verified installer provenance and teach users to reject instructions that require copying unfamiliar commands into a system shell. Correlate unusual library loading, script execution, and persistence changes rather than assessing each event in isolation. Isolate suspicious hosts, investigate outbound tunneling and directory reconnaissance, and confirm clean recovery options before returning systems to service.
Source: https://www.forescout.com/blog/from-clickfix-to-c2-comparative-analysis-of-two-malware-loaders/
10/14/26
KGB RAT combines Windows surveillance and credential theft
MEDIUM-HIGH
+ —
Intel Source:
Group-IB
Intel Name:
KGB RAT combines Windows surveillance and credential theft
Date of Scan:
10/14/26
Impact:
MEDIUM-HIGH
Summary:
Group-IB analyzed KGB RAT, a Windows remote-access trojan that combines surveillance, credential theft, and persistence under the guise of administration software. The examined samples use the Microsoft managed-code platform and share code with an established commodity trojan family. Observed capabilities include keylogging, screenshots, credential-vault access, clipboard theft, host reconnaissance, and collection of security-product information. Persistence through scheduled execution and startup settings allows continued access after the initial launch. Configurations reveal multiple operators, controller communication, and notification channels, including a case of financially motivated clipboard substitution. Initial execution still requires the victim to launch the delivered program. The seller advertises additional capabilities and strong evasion, but those claims do not establish that every advertised module operated in the analyzed samples. Behavioral evidence remains useful even when static obfuscation changes, and the report does not demonstrate broad enterprise victim scale. Defenders should prioritize application allowlisting and restrict unapproved remote administration software. Correlate credential access, surveillance behavior, and suspicious persistence through endpoint telemetry, investigate unusual outbound communications, and contain affected hosts. Review potentially exposed accounts before restoring trusted access.
Source: https://www.group-ib.com/blog/kgb-rat-windows-fud-trojan/
10/14/26
RuDNS fake utility delivers Arcane credential stealer
MEDIUM-HIGH
+ —
Intel Source:
Varist
Intel Name:
RuDNS fake utility delivers Arcane credential stealer
Date of Scan:
10/14/26
Impact:
MEDIUM-HIGH
Summary:
Varist analyzed a fake Windows DNS utility that delivers Arcane Stealer and exposes browser sessions and application credentials to theft. The installer provides no genuine DNS-management functionality and instead starts a layered execution chain through installer actions and scripts. It seeks elevated access when necessary and adds its installation location to antivirus exclusions, reducing protection against subsequent stages. Successive encrypted payloads use in-memory loading, while an argument check presents a decoy when a stage launches outside the expected sequence. The final stealer collects browser passwords, cookies, autofill and payment data, messaging tokens, and file-transfer application information. It also profiles the host, captures screenshots, and uploads stolen information to remote collection infrastructure. The analysis describes techniques for accessing protected browser data but does not establish campaign scale or a verified victim geography. Defenders should prioritize approved installers and validate software provenance before allowing execution. Investigate unexpected antivirus exclusions and installer-to-script process chains, including elevation attempts and in-memory loading. Contain affected endpoints, assess suspicious outbound uploads, and revoke exposed sessions and credentials across browser, messaging, and file-transfer accounts.
Source: https://www.varist.com/blogs-news/Unknown-Threats-Digest-RuDNS-A-Fake-DNS-Utility-Delivering-Arcane-Stealer
10/14/26
Fake Calendly recruitment pages relay account authentication
MEDIUM-HIGH
+ —
Intel Source:
Malwarebytes
Intel Name:
Fake Calendly recruitment pages relay account authentication
Date of Scan:
10/14/26
Impact:
MEDIUM-HIGH
Summary:
Malwarebytes identified fake interview-booking pages that impersonate Meta recruitment and steal account credentials and authentication responses, creating risks for business accounts and email access. The campaign imitates Calendly scheduling, collects contact information, and uses a false verification step before presenting a simulated Facebook sign-in window. It then requests security codes and follows with a Google authentication sequence that can collect passwords, recovery information, or approval responses. An attacker-controlled backend directs the interaction in real time, allowing stolen credentials and multifactor responses to be used promptly. Related sites share reusable kit features that support repeated deployment, but the report does not establish a victim count or targeting geography. Compromised Facebook access can expose business pages and advertising accounts, while Google email access can reveal sensitive information and support password resets elsewhere. Defenders should prioritize independent verification of recruiters and vacancies and inspect the genuine browser address rather than trusting an embedded sign-in window. Use passkeys or security keys and reject unexpected authentication prompts. After exposure, change passwords from a trusted device, revoke sessions, review recovery and forwarding settings, and check business-page and advertising activity.
Source: https://www.malwarebytes.com/blog/threat-intel/2026/10/fake-calendly-meta-careers-pages-steal-facebook-and-google-logins
10/13/26
Ivanti Connect Secure Chained Zero-Day Exploitation
HIGH
+ —
Intel Source:
Volexity
Intel Name:
Ivanti Connect Secure Chained Zero-Day Exploitation
Date of Scan:
10/13/26
Impact:
HIGH
Summary:
Volexity documented active exploitation of an Ivanti Connect Secure vulnerability chain that gives unauthenticated attackers root-level command execution on enterprise VPN appliances and access to credentials used inside the network. The investigation began after suspicious activity was detected at a monitored customer whose appliance was running the latest available firmware at the time of compromise. The chain combines authentication bypass and remote code execution, tracked together as CVE-2026-104935, with a separate privilege escalation flaw. The operator collected appliance configuration, domain credentials, Active Directory metadata, local administrator passwords, and BitLocker recovery keys, then carried out password spraying and network reconnaissance. Commands and results moved through repeated exploitation requests rather than a persistent backdoor or web shell. Selective log deletion and temporary memory changes reduced forensic visibility, and the vendor integrity checker could miss the compromise. Volexity disrupted the observed operation, but the attacker's ultimate objective remains unknown. Defenders should prioritize vendor updates for affected appliances and track the pending privilege escalation fix. Investigate suspicious devices with network and memory forensics, preserve evidence, and review activity originating from the VPN appliance. Reset exposed directory and service-account credentials, investigate possible lateral movement, and reestablish trust in the remote-access environment before returning systems to service.
Source: https://www.volexity.com/blog/2026/10/13/a-series-of-unauthenticated-events-chained-exploitation-of-ivanti-connect-secure-zero-days/
10/13/26
Citrix NetScaler Exploitation Deploys Modified Platypus
HIGH
+ —
Intel Source:
Binary Defense
Intel Name:
Citrix NetScaler Exploitation Deploys Modified Platypus
Date of Scan:
10/13/26
Impact:
HIGH
Summary:
Binary Defense investigated exploitation of Citrix NetScaler appliances that delivered a modified Platypus remote administration tool, creating a path to remote command execution and persistent access on enterprise gateways. The reported activity involves CVE-2026-88771, a pre-authentication command injection flaw, and CVE-2026-88772, a memory overflow affecting the datagram-based secure transport service. Investigators examined two potentially compromised appliances and recovered a FreeBSD implant from one, while related analysis identified a Windows agent from the same private tool fork. The agents support encrypted enrollment, interactive shell access, file operations, plugins, and updates, with additional modules for privilege escalation, persistence, timestamp manipulation, and cleanup. Those capabilities should not be treated as confirmed outcomes: the recovered appliance evidence did not establish completed installation or successful controller enrollment. The report also describes logging gaps that can hide exploitation from centralized monitoring and notes that failover and configuration synchronization can complicate investigations of paired appliances. Defenders should apply the vendor security updates to both active and standby nodes and reduce unnecessary gateway exposure. Preserve and review local web, shell, system, and authentication logs on each node, correlate unusual outbound traffic, and verify centralized log coverage. Investigate unauthorized binaries, scheduled execution, configuration changes, and administrative access, then remove persistence from both nodes before returning the pair to service.
Source: https://binarydefense.com/resources/blog/platypwned-when-a-citrix-0-day-lays-an-egg
10/13/26
MALFEX Native npm Loader Targets Windows Developers
MEDIUM-HIGH
+ —
Intel Source:
SafeDep
Intel Name:
MALFEX Native npm Loader Targets Windows Developers
Date of Scan:
10/13/26
Impact:
MEDIUM-HIGH
Summary:
SafeDep identified a MALFEX campaign evolution that moves malicious npm loader logic into a compiled Windows addon, exposing developer workstations to in-memory payload execution when an application loads the package. The operator replaced removed JavaScript loaders with a native version that downloads encrypted shellcode, decrypts it, and injects it into a Windows system process. Execution does not depend on an installation script, so controls that only block package installation hooks do not prevent this path. The addon checks for an interactive Windows console, which can suppress activity in automated testing and build environments that redirect output. SafeDep also recovered a related downloader and a separate stealer and remote-access payload with credential collection, persistence, and log-tampering capabilities, but did not confirm that the native loader delivers that same final payload. Public dependency searches found no declared consumers, and download counts did not establish how many workstations were compromised. Defenders should review package provenance, dependency records, and native addons across Windows development environments, then remove unapproved campaign-related dependencies. Isolate potentially affected workstations, preserve memory, and investigate abnormal child processes, persistence, and inconsistencies between local and centralized logs. Rotate potentially exposed development, cloud, source-control, and browser credentials from trusted devices, and restore affected workstations from trusted sources.
Source: https://safedep.io/malfex-imgdecode-npm-native-addon-loader/
10/12/26
Apex Accelerator Malicious npm Campaign
MEDIUM-HIGH
+ —
Intel Source:
Elastic Security Labs
Intel Name:
Apex Accelerator Malicious npm Campaign
Date of Scan:
10/12/26
Impact:
MEDIUM-HIGH
Summary:
Elastic Security Labs reported a campaign that posed as a Web3 accelerator and distributed malicious developer tools through npm packages, backdoored repositories, and a fake audit service, creating supply-chain and identity risk for start-ups and engineering teams. The operation used a legitimate-looking coding assistant and security-audit workflow to deliver macOS and Windows malware, execute attacker-selected commands, collect host and browser data, and route AI coding traffic through attacker-controlled services. Elastic also found deceptive audit output and persistent registration of a remote AI tool service, creating a risk that prompts, code snippets, tool output, and credentials could be exposed or manipulated. Block the identified malicious packages and repositories, review package-lock files and build records, and restrict developer installations to approved package sources. Investigate developer endpoints for unexpected AI tool configuration changes, unusual child processes, and suspicious data transfer. Rotate exposed development, cloud, and source-control credentials, and rebuild affected systems from trusted images.
Source: https://www.elastic.co/security-labs/threat-command/apex-accelerator-malicious-npm-packages
10/12/26
AtlasRAT C2 Detection Update
MEDIUM
+ —
Intel Source:
NETRESEC
Intel Name:
AtlasRAT C2 Detection Update
Date of Scan:
10/12/26
Impact:
MEDIUM
Summary:
NETRESEC documented a newer AtlasRAT command-and-control protocol variant that changes earlier network behavior and can reduce the effectiveness of detections built only for legacy traffic. The research describes a progression from earlier clear-text behavior to encrypted communications and identifies shared infrastructure observations that suggest operator overlap with related remote-access activity, while noting that the specific attribution remains circumstantial. The report provides network-focused detection material, including Suricata signatures and TLS characteristics, that can help defenders recognize both older and newer protocol variants. Deploy and test the published network detections in monitored environments, and hunt historical network telemetry for matching protocol behavior and unusual outbound encrypted sessions. Investigate affected endpoints for unauthorized remote-access execution, persistence, and credential-access activity. Segment high-value systems, restrict unnecessary outbound connectivity, and retain endpoint and network telemetry long enough to support retrospective hunting. Treat protocol changes as a trigger to validate detection coverage after each intelligence update.
Source: https://www.netresec.com/?page=Blog&month=2026-10&post=Detecting-AtlasRAT-C2
10/12/26
ICMacOS Persistent RAT Service
MEDIUM-HIGH
+ —
Intel Source:
SOCRadar
Intel Name:
ICMacOS Persistent RAT Service
Date of Scan:
10/12/26
Impact:
MEDIUM-HIGH
Summary:
SOCRadar reported ICMacOS, a macOS malware-as-a-service offering marketed since September 2026 that combines information theft with persistent remote-access capabilities and a companion social-engineering delivery panel. The research states that the service targets macOS environments and can collect browser, wallet, credential, and host data while using persistence and command-execution features that raise risk for organizations with developer, executive, or cryptocurrency-related Mac fleets. Its distribution model lowers the barrier for multiple operators to deploy the tooling, increasing the chance of varied campaigns and changing delivery methods. Review macOS endpoint telemetry for unexpected terminal activity, new login items, unauthorized launch agents, and unapproved remote-access software. Require managed application installation, enforce least privilege, and limit administrator approval paths for developer and executive devices. Audit browser and credential-store access, rotate credentials where compromise is suspected, and isolate then rebuild systems that show confirmed unauthorized persistence. Preserve forensic images and relevant logs before remediation when policy permits.
Source: https://socradar.io/blog/icmacos-infostealer-rat/
10/12/26
AhsayCBS Active Exploitation Chain
HIGH
+ —
Intel Source:
Rescana
Intel Name:
AhsayCBS Active Exploitation Chain
Date of Scan:
10/12/26
Impact:
HIGH
Summary:
Rescana reported active exploitation of a vulnerability chain in AhsayCBS Backup Console that can progress from authentication bypass to remote code execution, with observed follow-on web shells and cryptomining activity. The report warns that the reported fixed release did not fully close the issue, making externally reachable backup-management infrastructure a high-priority enterprise exposure. Rescana describes attacker actions that can establish persistence, execute commands, and consume system resources, creating operational and recovery risk for organizations that rely on affected backup consoles. Identify every AhsayCBS deployment, remove public exposure where possible, and apply the vendor mitigation or a validated compensating control rather than relying only on the cited release. Review console, web-server, endpoint, and network logs for unauthorized authentication, command execution, new services, or unexpected resource use. Isolate confirmed affected hosts, remove persistence, rotate credentials used by the platform, and verify that backups and recovery procedures remain trustworthy.
Source: https://www.rescana.com/post/ahsaycbs-cve-2026-105133-105134-exploited-backup-console-rce
10/9/26
GhostAction GitHub Credential Theft
HIGH
+ —
Intel Source:
StepSecurity
Intel Name:
GhostAction GitHub Credential Theft
Date of Scan:
10/9/26
Impact:
HIGH
Summary:
StepSecurity reports that GhostAction attackers are compromising GitHub maintainer accounts to steal repository secrets and credentials from full commit histories, creating immediate risk to enterprise development environments and connected cloud services. The researchers analyzed malicious workflows, commit metadata, and execution logs from an October 8 campaign that affected 345 repositories through two compromised accounts. The injected workflows masqueraded as security improvements and ran under legitimate maintainer identities, allowing changes to reach default branches without normal review. Collection combined configured Actions secrets with credentials found across the working tree and historical commits, including cloud, AI service, and SaaS access material. StepSecurity confirmed successful exfiltration from an enterprise-owned repository and found hundreds of repositories that still contained malicious workflows. At publication, the researchers had not observed malicious package releases using the exposed publishing credentials. Defenders should revoke compromised maintainer access, rotate repository, publishing, and historical credentials reachable by affected workflows, and remove malicious workflows from every branch and affected fork. Review execution logs and suspend releases until the environment is clean, then require review for workflow changes, enforce execution approvals where feasible, and restrict runner network access to approved destinations.
Source: https://www.stepsecurity.io/blog/ghostaction-returns
10/9/26
PaperCut Authentication Bypass and RCE Chain
HIGH
+ —
Intel Source:
watchTowr Labs
Intel Name:
PaperCut Authentication Bypass and RCE Chain
Date of Scan:
10/9/26
Impact:
HIGH
Summary:
watchTowr Labs documents authentication bypasses and remote code execution in PaperCut that defeat earlier fixes, leaving incompletely updated print-management servers exposed to compromise. The researchers reproduced a chain against PaperCut NG 26.0.4-PO build 76508 that allows an unauthenticated attacker to reset administrative access and then execute commands through scan-to-fax functionality. The authentication weakness stems from incomplete authorization checks across setup forms, while the execution flaw allows administrator-controlled fax settings to reach an operating-system command processor. The report distinguishes these researcher-demonstrated bypasses from an earlier chain that attackers exploited in the wild against version 26.0.3. It also documents successive fixes, with the final code-execution correction released in version 26.0.5, and does not establish malicious exploitation of the newly described bypass chain. Organizations should verify installed builds, apply the release that fixes the complete chain, and limit management-interface exposure to trusted networks. Investigate unexpected administrator-password changes, setup activity, fax configuration changes, and unusual child processes from the print service. If compromise is confirmed, isolate affected servers, preserve relevant logs, rotate exposed administrative credentials, and rebuild from trusted sources before restoring service.
Source: https://labs.watchtowr.com/death-by-a-thousand-papercuts-papercut-pre-auth-rce-chain-and-patch-bypasses-wt-2026-0141-0144-cve-2026-82077-cve-2026-82078-cve-2026-81578/
10/9/26
VorishkaStealer Windows Data Theft
MEDIUM
+ —
Intel Source:
CYFIRMA
Intel Name:
VorishkaStealer Windows Data Theft
Date of Scan:
10/9/26
Impact:
MEDIUM
Summary:
CYFIRMA reports VorishkaStealer activity targeting Windows systems, where theft of credentials, local files, and clipboard contents can expose business information and support further account compromise. Its October 9 analysis describes anti-analysis checks at execution and inspection of registry configuration, local files, processes, and system state. The sample monitors copied content and collects documents, configuration data, and other sensitive material from the workstation. These capabilities place information used during normal employee activity at risk, including passwords and confidential text copied between applications. CYFIRMA identifies global targeting, but observed sessions showed limited outward traffic and did not establish durable persistence, so the report does not demonstrate a broad enterprise outbreak or confirmed large-scale exfiltration. Defenders should maintain endpoint protection and monitor unusual process trees, command-shell activity, and unauthorized security-setting changes on Windows devices. Confirmed infections should be isolated, investigation data preserved, startup locations reviewed, and affected systems restored only after cleanup. Rotate credentials used on compromised endpoints, verify multifactor authentication enrollment, and review account access to limit reuse of exposed authentication material.
Source: https://www.cyfirma.com/news/weekly-intelligence-report-9-oct-2026/
10/8/26
UAT-11985 AI-Assisted Google AitM Phishing
MEDIUM-HIGH
+ —
Intel Source:
Cisco Talos
Intel Name:
UAT-11985 AI-Assisted Google AitM Phishing
Date of Scan:
10/8/26
Impact:
MEDIUM-HIGH
Summary:
Cisco Talos reports that UAT-11985 used AI-assisted event lures to deliver real-time adversary-in-the-middle phishing against Google accounts, creating a direct risk of credential theft, MFA interception, and session takeover. The campaign targeted research organizations in Taiwan and used convincing event-themed content to steer recipients into a live proxy workflow that relayed authentication activity to the attackers. Talos assesses the activity as phishing that can bypass protections based only on stolen passwords because the workflow captures authentication data during an active login session. The report provides publisher-controlled detection material for mail and network defenses, allowing security teams to identify related delivery and proxy behavior without relying on third-party reporting. Defenders should prioritize likely research and event-facing populations with targeted awareness, enforce phishing-resistant MFA, and require reauthentication for high-risk sessions. Email, web, DNS, and identity telemetry should be tuned for deceptive event invitations, unusual sign-in flows, and token reuse. Suspicious account sessions should be investigated promptly, affected sessions revoked, exposed credentials rotated, and conditional access policies reviewed for gaps.
Source: https://blog.talosintelligence.com/uat-11985/
10/8/26
Hikvision DVR Exploitation Attempts in Ukraine
MEDIUM-HIGH
+ —
Intel Source:
GreyNoise
Intel Name:
Hikvision DVR Exploitation Attempts in Ukraine
Date of Scan:
10/8/26
Impact:
MEDIUM-HIGH
Summary:
GreyNoise observed a concentrated increase in attempts to exploit exposed Hikvision digital video recorders in Ukraine, showing that known device weaknesses remain useful to operators seeking access to edge infrastructure during a period of heightened regional pressure. The activity began on September 21, persisted for nine days, and then stopped, with the observed traffic concentrated in a small set of sources. GreyNoise does not claim a confirmed link between the scanning activity and concurrent military events, but the timing and focus warrant defensive attention for organizations operating video, physical security, or connected-device estates in Ukraine. The primary report includes concrete publisher-observed artifacts and a detection reference that support direct validation. Organizations should identify all exposed recorders and management interfaces, remove unnecessary Internet exposure, and apply available vendor fixes or compensating controls. Administrative access should be restricted through segmented networks and strong authentication, with device logs, firewall events, and authentication records reviewed for suspicious access. Compromised devices should be rebuilt from trusted images where feasible, and teams should validate backups, monitoring coverage, and recovery procedures for physical-security systems.
Source: https://www.greynoise.io/blog/hikvision-camera-exploitation-attempts-ukraine
10/8/26
Novinarya Android Financial Stealer
MEDIUM-HIGH
+ —
Intel Source:
STAR Labs
Intel Name:
Novinarya Android Financial Stealer
Date of Scan:
10/8/26
Impact:
MEDIUM-HIGH
Summary:
STAR Labs reports on Novinarya, an Android banking and cryptocurrency stealer that hides its active command infrastructure behind encrypted data and a legitimate marketplace profile, complicating routine network blocking and static analysis. The malware targets 81 financial applications in Iran, including banking and cryptocurrency services, and can collect credentials through phishing WebViews, intercept SMS one-time passcodes, and abuse notification access. Its two-layer packaging and rotating infrastructure design increase the risk that mobile users and financial-service customers will install a loader that appears limited before it retrieves the theft capability. The primary report provides concrete, accessible artifacts that support threat-specific detection and validation. Defenders should block unapproved application sources, enforce managed-device policy, and restrict high-risk accessibility, notification, and package-install permissions. Mobile authentication and transaction telemetry should be monitored for unusual device registration, SMS interception patterns, and account access anomalies. Where exposure is suspected, affected users should be notified, credentials and sessions reset, and fraud monitoring coordinated with incident response teams.
Source: https://starlabs.sg/blog/2026/10-novinarya-an-android-stealer-that-hides-its-live-c2-in-a-shop-bio/
10/6/26
CyberXero Enterprise Intrusions and Access Brokering
HIGH
+ —
Intel Source:
SOCRadar
Intel Name:
CyberXero Enterprise Intrusions and Access Brokering
Date of Scan:
10/6/26
Impact:
HIGH
Summary:
SOCRadar reports that CyberXero is conducting active enterprise intrusions through automated web exploitation and directed attacks on Ukrainian critical infrastructure, creating risks of access resale and sensitive data theft. The campaign combined broad activity against WordPress and e-commerce systems with targeted reconnaissance and manual exploitation of Ukrainian energy and utilities organizations. Recovered working directories, session logs, and victim exports linked eight infrastructure nodes and showed confirmed exfiltration from four Ukrainian organizations, including records concerning more than 628,000 individuals. In one automated execution, the operator scanned 4,708 targets, confirmed access to 429 administration panels, and deployed 32 web shells within 61 seconds. CyberXero also used AI agents for reconnaissance, exploitation, and post-compromise tasks alongside conventional offensive tooling; recovered logs showed refusals and repeated attempts to reframe requests. The research did not prove access sales to a particular buyer, and reconnaissance activity should not be treated as confirmed compromise. Defenders should prioritize patching exposed web applications and restricting database access, then audit privileged accounts, plugins, remote access, and unexpected data exports. AI agent logs should also be protected with access controls, encryption, retention policies, and operational auditing.
Source: https://socradar.io/blog/cyberxero-ai-iab-ukraine-critical-infrastructure/
10/6/26
Akira Ransomware Intrusion with Persistent Tunneling
HIGH
+ —
Intel Source:
Huntress
Intel Name:
Akira Ransomware Intrusion with Persistent Tunneling
Date of Scan:
10/6/26
Impact:
HIGH
Summary:
Huntress reconstructed an Akira ransomware intrusion that combined remote access, security control impairment, data exfiltration, and file encryption, showing how quickly disruption can escalate even when early endpoint telemetry is missing. In the investigated organization, the endpoint agent was installed after compromise, which limited visibility into initial access and earlier attacker activity. Windows event logs showed access through Remote Desktop Protocol from a workstation outside the customer environment, followed by access to the antivirus console and stoppage of associated services. The attacker also used a process dumping utility, apparently for credential theft, and deployed GOST tunneling and Rclone tooling. Registry artifacts, ransomware logs, and PowerShell events established repeated encryption activity against shared folders and deletion of volume shadow copies. The tunneling component was deployed after encryption had already begun and provided a potential persistence mechanism. These findings describe one investigated incident rather than a measured increase in campaign prevalence. Defenders should reduce remote access exposure and enforce MFA, correlate authentication, service changes, suspicious execution, and file activity across endpoint and Windows logs, and investigate residual tunneling access while validating protected backups and recovery readiness.
Source: https://www.huntress.com/blog/mapping-akira-ransomware-attack
10/6/26
Atlassian Pre-Authentication File Read and Identity Risk
HIGH
+ —
Intel Source:
watchTowr
Intel Name:
Atlassian Pre-Authentication File Read and Identity Risk
Date of Scan:
10/6/26
Impact:
HIGH
Summary:
watchTowr demonstrated a pre-authentication file-read vulnerability across multiple Atlassian server products that can expose application secrets and create identity compromise risk in certain deployments. The research covers CVE-2026-21589 and verified the shared resource-handling weakness against Jira, Confluence, and Bitbucket. The vulnerable routing logic translates unusual path separators and permits traversal into files within the application server context. Researchers did not demonstrate traversal outside that context, so the issue should not be described as unrestricted host filesystem access. In a Jira deployment integrated with Crowd, readable configuration contained the application credentials used to communicate with the centralized identity service. Where Crowd was reachable and accepted those credentials, the researchers demonstrated user creation and privilege modification. Restrictive network allowlists could make that chain harder to exploit, and the report does not establish active exploitation in production. Organizations should prioritize vendor-fixed releases across affected Atlassian deployments, reduce public exposure, restrict access to Crowd and related identity services, review resource requests and account or privilege changes, and assess whether exposed application credentials require rotation.
Source: https://labs.watchtowr.com/you-wont-hear-about-these-even-in-myths-atlassian-jira-confluence-and-more-pre-auth-arbitrary-file-read-cve-2026-21589/
10/6/26
Blinder Tunnel Recruitment Lures Target Critical Infrastructure
MEDIUM-HIGH
+ —
Intel Source:
Palo Alto Networks
Intel Name:
Blinder Tunnel Recruitment Lures Target Critical Infrastructure
Date of Scan:
10/6/26
Impact:
MEDIUM-HIGH
Summary:
Unit 42 reports that an Iranian state-aligned activity cluster used fraudulent recruitment material to compromise developer environments supporting critical infrastructure, giving attackers covert access and network tunneling capabilities. The Blinder Tunnel campaign targeted Iraqi organizations in March 2026, following infrastructure preparation observed as early as November 2025. The attackers impersonated Dubai Airports recruiters and delivered a personalized Visual Studio coding challenge; the report found no evidence that Dubai Airports systems were compromised. The infection chain abused project build behavior, application domain management, and DLL sideloading to execute custom loaders, remote access payloads, and tunneling components. GitHub infrastructure supplied keys, payloads, and fallback routing information, helping malicious communications resemble legitimate developer activity, and GitHub later removed the infrastructure identified in the investigation. Related findings linked the cluster to credential harvesting against an Israeli entity and a wider regional targeting profile, but did not prove attribution to a specific established Iranian group. Defenders should verify recruitment and project provenance before builds, monitor developer processes, persistence changes, and unusual cloud platform communications, and investigate affected endpoints and identities. Unauthorized tunneling should be contained and credentials rotated where exposure is established.
Source: https://unit42.paloaltonetworks.com/blinder-tunnel-targets-critical-infrastructure/
10/6/26
Stolen AWS Credentials Tested for Amazon Bedrock Access
MEDIUM-HIGH
+ —
Intel Source:
Datadog
Intel Name:
Stolen AWS Credentials Tested for Amazon Bedrock Access
Date of Scan:
10/6/26
Impact:
MEDIUM-HIGH
Summary:
Datadog reports that attackers are testing stolen AWS credentials for Amazon Bedrock access, providing an early warning of unauthorized AI consumption and potential financial loss. Researchers observed credential harvesting infrastructure probing more than 80 customers and separately identified two tools that validate identities, enumerate available models across regions, and attempt minimal model requests. One tool also checks promotional credits, suggesting interest in the financial value of usable access. The examined platform distinguishes Bedrock-capable credentials and handles service-specific tokens, although researchers could not observe its backend validation or establish associated AWS activity. Across the previous 30 days, telemetry from 12 organizations showed similar suspicious model discovery behavior, with one case progressing to repeated invocation attempts that received access-denied responses. Datadog could not directly link the analyzed tools to those telemetry events, so successful resource abuse should not be assumed. Defenders should remove exposed credentials, restrict long-term key permissions, and correlate identity validation, cross-region model enumeration, and unexpected invocation with each identity's normal usage. Anomalous Bedrock access should be investigated promptly, affected identities contained, and potential unauthorized consumption reviewed.
Source: https://securitylabs.datadoghq.com/articles/beyond-valid-credentials-how-exposed-aws-keys-are-tested-for-amazon-bedrock-access/
10/6/26
IronChain Destructive Ransomware Build
MEDIUM-HIGH
+ —
Intel Source:
ANYRUN FZCO
Intel Name:
IronChain Destructive Ransomware Build
Date of Scan:
10/6/26
Impact:
MEDIUM-HIGH
Summary:
ANY.RUN analyzed a new IronChain build that combines destructive file handling with privileged persistence and attempted security impairment, making reliable recovery uncertain if execution succeeds. The Windows payload appeared in public telemetry in September and packaged four kernel drivers alongside an extensive attack workflow. Static analysis found one coherent driver interface for process termination, while the other three integrations were broken or unused. Existing sandbox telemetry showed early persistence and service-suppression attempts, but the research did not prove driver loading, process termination, or successful raw disk writes. Programming mistakes also narrowed ordinary file traversal, while discarded cryptographic material prevented a dependable reversal of affected encryption. The report identified no verified victim, working command server, zero-day, or named actor, and shared network services should not be treated as malicious solely because they appeared with this sample. Defenders should enforce vulnerable-driver controls and restrict untrusted execution, then investigate combinations of new privileged tasks, driver services, defense impairment, and destructive file activity. Suspected systems should be isolated with evidence preserved, boot-disk damage assessed before rebooting, and recovery prepared from trusted media and tested isolated backups.
Source: https://any.run/cybersecurity-blog/ironchain-analysis/
10/6/26
Phishing Installs Action1 for Unauthorized Remote Access
MEDIUM
+ —
Intel Source:
SANS Institute
Intel Name:
Phishing Installs Action1 for Unauthorized Remote Access
Date of Scan:
10/6/26
Impact:
MEDIUM
Summary:
SANS Internet Storm Center reports on a phishing delivery chain that installs the legitimate Action1 remote management agent, allowing an attacker to establish persistent remote access through trusted software and its cloud infrastructure. The examined email carried a fake invoice PDF that redirected the recipient to a script. That script displayed a decoy document while downloading and installing an agent package whose components were not flagged as malicious by the public scanning service referenced in the report. The installed software used a persistent Windows service and stored customer configuration and key material locally. The researcher assessed that the attacker was abusing the vendor's cloud infrastructure, potentially through a free or trial account, rather than compromising the vendor itself. A second sample used a shipping-document theme and a different script delivery route to install the same package. The report covers two observed samples and does not provide a victim count or demonstrate subsequent ransomware deployment. Defenders should block unsolicited script execution, validate remote management installer provenance, audit newly installed agents, services, and tenant ownership, and review phishing exposure and unauthorized remote sessions. Unapproved access should be removed after preserving relevant logs.
Source: https://dshield.org/diary/More+RMM+Tools+In+the+Wild/33400/
10/6/26
RevStealer Uses Weighted Sandbox Evasion
MEDIUM
+ —
Intel Source:
Joe Security LLC
Intel Name:
RevStealer Uses Weighted Sandbox Evasion
Date of Scan:
10/6/26
Impact:
MEDIUM
Summary:
Joe Security reports that RevStealer can suppress its main functionality when it recognizes an analysis environment, making a quiet automated examination an unreliable basis for clearing suspicious enterprise files. The analyzed sample's final payload combined nine weighted environmental checks with a direct host identity gate. Those checks evaluate graphics and display characteristics, processor topology, memory, running processes, uptime, timing behavior, and hardware consistency. Once the accumulated score reaches its threshold, the payload resets state and terminates after a short randomized delay, allowing several weak signals or one strong signal to stop execution. A hardware video encoder probe was particularly significant because many virtual machines lack the relevant capability and some analysis systems do not monitor the associated API calls. Researchers supported attribution through matching code structures and infrastructure across investigated and reference payloads. The report does not establish delivery scale or a specific victim geography. Defenders should retain suspicious files for deeper specialist review when automated behavior is limited, correlate endpoint and network evidence with static analysis and environment coverage, and investigate confirmed infections for credential exposure. Affected systems should be contained and compromised secrets rotated as warranted.
Source: https://joesecurity.org/blog/6469689575970038406
10/5/26
FortiMail Active Exploitation
HIGH
+ —
Intel Source:
Fortinet
Intel Name:
FortiMail Active Exploitation
Date of Scan:
10/5/26
Impact:
HIGH
Summary:
Fortinet reports active exploitation of an unauthenticated FortiMail vulnerability that permits arbitrary file writes and exposes enterprise email gateways to unauthorized code or command execution. The flaw combines path traversal with improper handling of NULL bytes in the appliance web interface, which the vendor identifies as the affected component. Affected releases span the FortiMail 8.0, 7.6, 7.4, and 7.2 branches, making version verification important across mixed appliance estates. Fortinet marks the issue as known exploited and urges customers to apply its workaround. An October 5 solution update adds fixed-version guidance, giving defenders a supported remediation path alongside temporary exposure reduction. The advisory does not establish the scale of exploitation or identify affected countries, so the confirmed activity warrants urgent action without assuming a worldwide incident. Defenders should upgrade each affected branch to the corrected release, moving older installations to a supported corrected branch where required. Until upgrades are complete, disable Identity-Based Encryption or restrict public webmail access to trusted private networks using the vendor's guidance. Review appliance system-event and encryption logs for the documented attack behavior, preserve relevant records, and investigate matching activity.
Source: https://fortiguard.fortinet.com/psirt/FG-IR-26-175
10/5/26
SubQuery Supply Chain Credential Theft
HIGH
+ —
Intel Source:
StepSecurity
Intel Name:
SubQuery Supply Chain Credential Theft
Date of Scan:
10/5/26
Impact:
HIGH
Summary:
StepSecurity identified a malicious release in the SubQuery dependency ecosystem that can expose developer workstations, continuous integration runners, and accessible cloud services to credential theft and remote shell access. The payload starts during installation and package import, so disabling installation scripts alone does not prevent execution. Researchers traced changes that replaced built package contents with an external artifact immediately before publication, while an existing shared dependency relationship could pull the affected release into downstream applications. Of 77 reviewed packages, 19 latest versions had a potential dependency path to the affected library. The decoded payload collects local secrets and process environment data, attempts permission-dependent access to cloud and cluster secrets, and can insert workflows intended to extract repository secrets. These capabilities do not establish successful theft from a particular victim, and available permissions and implementation defects can limit execution. Defenders should exclude the affected release from direct and transitive dependency resolution, then validate installed trees and lockfiles. If installation or import could have executed the payload, isolate hosts and runners, preserve process and network evidence, and rebuild from trusted images. Revoke and rotate accessible credentials, then audit repository workflow changes and cloud secret access for unauthorized activity.
Source: https://www.stepsecurity.io/blog/subql-ecosystem-compromised
10/5/26
ClingSTUN Linux Device Backdoor
HIGH
+ —
Intel Source:
Fortinet
Intel Name:
ClingSTUN Linux Device Backdoor
Date of Scan:
10/5/26
Impact:
HIGH
Summary:
Fortinet reports an evolving ClingSTUN campaign that exploits unpatched Internet-facing devices and turns compromised Linux systems into persistent proxy nodes under remote control. Observed access methods span multiple device vendors and include command injection, authentication bypass, and buffer overflow vulnerabilities, creating exposure across network, security, and IoT infrastructure. The malware supports remote command execution and contains exploits for self-propagation, although the report does not establish mass exploitation or a confirmed victim count. Its persistence and concealment mechanisms alter boot behavior, manipulate watchdog functions, terminate competing processes, and obscure process information. ClingSTUN also uses legitimate public STUN services to discover external address mappings and maintain connectivity, making that traffic difficult to distinguish from normal communications without host context. The report does not verify how the operator obtains mappings or delivers control traffic through address translation. Defenders should inventory exposed devices, patch actively exploited flaws, replace or isolate unsupported firmware, and restrict unnecessary public services. Correlate unusual UDP activity and recurring connections with suspicious processes, boot changes, and unexplained process termination rather than classifying public STUN services as malicious by default. Investigate affected devices and restore compromised systems with trusted firmware, confirming that unauthorized persistence has been removed.
Source: https://www.fortinet.com/blog/threat-research/clingstun-linux-backdoor-abuses-public-stun-infrastructure
10/5/26
RubyGems Developer Supply Chain Compromise
MEDIUM-HIGH
+ —
Intel Source:
OpenSourceMalware
Intel Name:
RubyGems Developer Supply Chain Compromise
Date of Scan:
10/5/26
Impact:
MEDIUM-HIGH
Summary:
OpenSourceMalware identified 42 malicious RubyGems packages targeting cryptocurrency and Web3 developers, creating software supply-chain risk for organizations whose developers install unfamiliar dependencies. The packages use typosquatting, borrowed project branding, and plausible utility names, with malicious native-extension installation logic that checks for a real developer workstation and delays activation for approximately 20 to 40 minutes. Eleven packages provide an interactive reverse shell, while another variant retrieves a cryptocurrency theft toolkit disguised as a wallet-protection utility. The toolkit attempts to install a trusted root certificate, alter proxy settings, and establish desktop autostart persistence. It intercepts browser traffic, replaces withdrawal destinations while preserving the visible original value, monitors clipboard contents, and searches for wallet data, seed phrases, and private keys. This can expose both transaction funds and reusable secrets, although the report does not establish a confirmed victim count or broad enterprise compromise. Defenders should verify dependency provenance and review recent installations for suspicious native-extension behavior. Audit affected endpoints for unauthorized certificate trust, proxy changes, persistence, and unusual child processes. If exposure is confirmed, isolate and rebuild compromised workstations, then use a clean device to replace exposed wallet material and rotate exchange credentials, API keys, and authentication factors.
Source: https://opensourcemalware.com/blog/malicious-crypto-shell
10/2/26
FortiMail Zero-Day Exploitation
HIGH
+ —
Intel Source:
Fortinet
Intel Name:
FortiMail Zero-Day Exploitation
Date of Scan:
10/2/26
Impact:
HIGH
Summary:
Fortinet reports active exploitation of an unauthenticated FortiMail vulnerability that allows arbitrary file writes and code or command execution on enterprise email gateways. The flaw combines path traversal with improper handling of null characters in the web interface. Affected releases include the 7.2, 7.4, 7.6, and 8.0 branches, with fixed releases forthcoming for supported branches. The advisory lists signs of compromise, including unexpected scheduled execution, abnormal administrator activity, encryption errors, and unauthorized remote email archiving settings. Check appliance integrity and configuration as well as reducing exposure. Fortinet confirms exploitation in the wild but does not identify an actor, victim locations, or the number of compromised organizations. Disable Identity-Based Encryption support or restrict webmail access to trusted private networks using the vendor's workarounds. Track the specified fixed releases and upgrade the appropriate branch when available. Review system and encryption logs, administrator changes, and remote archive settings for unauthorized activity, and refer suspicious appliances for incident investigation.
Source: https://fortiguard.fortinet.com/psirt/FG-IR-26-175
10/2/26
Check Point Management Server Root Compromise
HIGH
+ —
Intel Source:
Bishop Fox
Intel Name:
Check Point Management Server Root Compromise
Date of Scan:
10/2/26
Impact:
HIGH
Summary:
Bishop Fox demonstrates an unauthenticated exploit chain that grants root access to Check Point management servers, which control enterprise firewall policy and certificate trust. The chain combines a forged application login, an arbitrary file write, and scheduled execution. Researchers validated it on unpatched R81.10 and R82.10 laboratory systems. The report notes exploitation of the underlying vulnerability against a handful of customers and distinguishes those attacks from the laboratory findings. Compromise of Security Management or Multi-Domain Management servers can expose administrator credentials and the internal certificate authority. The tested chain bypasses the upgrade-service activity covered by the vendor's published log check, so a clean result does not rule out compromise. Install the applicable Jumbo Hotfix or security hotfix. Restrict management-service access to approved management stations through both network rules and the Trusted Clients configuration. Use file-integrity monitoring and execution telemetry to check for unexpected scheduled tasks, startup changes, and management processes launching shells. Investigate exposed, unpatched servers for possible intrusion and plan credential and certificate rotation as part of incident response.
Source: https://bishopfox.com/blog/weaponizing-check-point-management-cve-2026-93616
10/2/26
Warlock Ransomware Against Critical Infrastructure
HIGH
+ —
Intel Source:
Symantec
Intel Name:
Warlock Ransomware Against Critical Infrastructure
Date of Scan:
10/2/26
Impact:
HIGH
Summary:
Symantec reports ongoing Warlock ransomware intrusions against critical infrastructure and other organizations through exposed SharePoint servers. The Longlegs activity affected at least four organizations over the preceding two months: a water utility, telecommunications provider, regional government body, and university in Portuguese- and Spanish-speaking countries across Europe, Africa, and Latin America. The operators favor SharePoint vulnerabilities, deploy web shells, harvest application signing keys, and use forged application payloads for further execution. They also use library sideloading, legitimate remote-access tunneling, and vulnerable signed drivers to evade defenses. In one intrusion, tools used to disable security software reached at least 40 hosts within about two hours. Ransomware appeared on at least 33 hosts after staging through domain replication. Patch and restrict exposed SharePoint deployments, and investigate suspicious application execution or unauthorized changes affecting signing-key security. Audit privileged account changes, remote-access services, driver loading, and domain replication content for coordinated lateral movement. Contain confirmed intrusions across the domain, and verify isolated backups and recovery readiness before restoring affected services.
Source: https://www.security.com/threat-intelligence/warlock-ransomware-critical-infrastructure
10/2/26
GlassWorm-Linked Editor Extension Loaders
HIGH
+ —
Intel Source:
Socket
Intel Name:
GlassWorm-Linked Editor Extension Loaders
Date of Scan:
10/2/26
Impact:
HIGH
Summary:
Socket identifies a cluster of GlassWorm-associated editor themes on Visual Studio Marketplace and Open VSX that can deliver malicious code to developer environments. Researchers confirmed two malicious Marketplace builds and linked additional extensions through shared development history, publishing identities, and distinctive source-code features. One confirmed loader downloads and silently executes a Windows payload. Another decrypts embedded JavaScript and uses blockchain transaction data to locate additional code for execution in memory. Several related extensions were not weaponized in the versions examined, so their download counts do not indicate the number of infected systems. The broader cluster accumulated thousands of installations, and Marketplace removed reported extensions after notification. The code runs with the developer's available permissions, putting credentials and development resources at risk. Inventory extensions across both registries and compatible editors, remove confirmed malicious builds, and review related high-risk installations. Examine distributed extension packages and updates for unnecessary code execution, runtime decryption, network activity, and child processes. Public source repositories alone may not show these behaviors. Investigate follow-on execution on affected workstations and assess exposure of credentials and developer resources.
Source: https://socket.dev/blog/glassworm-vscode-themes
10/2/26
TIKTOUK Website and Cloud Credential Collection
HIGH
+ —
Intel Source:
LevelBlue
Intel Name:
TIKTOUK Website and Cloud Credential Collection
Date of Scan:
10/2/26
Impact:
HIGH
Summary:
LevelBlue describes TIKTOUK, a credential-collection toolkit that uses exposed WordPress data and client-delivered code to collect website, email, and cloud credentials. Its components probe websites, collect configuration and database option values, recover supported encrypted email settings with available keys, and scan JavaScript for secrets. Results are sent to a central service. Controlled testing confirmed these component behaviors but did not demonstrate successful exploitation of the referenced WordPress vulnerabilities or a live-site breach. Separate incident telemetry confirmed payload retrieval and controller communication. A leaked panel contained approximately 50,000 server-side credentials across about 37,000 domains, including hundreds of attacker-validated AWS keys. The credential exposure is substantial, though the simulated results do not confirm victim compromise. Remove public access to configuration files, repositories, backups, and debug data, and update affected WordPress installations. Use local web and endpoint logs to correlate unusual batch requests, changes in request encoding, sensitive-file access, and subsequent data submissions. Investigate exposed cloud and email credentials, rotate affected secrets, restrict their permissions, and audit associated services for unauthorized use.
Source: https://www.levelblue.com/blogs/spiderlabs-blog/tiktouk-tracing-a-wordpress-credential-collection-toolkit
10/2/26
Cling Botnet Abuses STUN Traffic
MEDIUM-HIGH
+ —
Intel Source:
Nozomi Networks
Intel Name:
Cling Botnet Abuses STUN Traffic
Date of Scan:
10/2/26
Impact:
MEDIUM-HIGH
Summary:
Nozomi Networks identifies Cling, an active botnet that compromises exposed networking and embedded devices and disguises command traffic as normal network traversal activity. Customer telemetry linked exploitation of a Realtek SDK flaw to malware that also includes scanning and exploit code for other router and recording-device vulnerabilities. The analyzed implant persists through startup configuration changes and replacement of a common download utility. It supports payload execution, propagation, tunneling, proxying, and denial-of-service attacks. Researchers observed commands over several days, including flooding tasks against a South Korean internet provider and a university cluster in the United States. The report confirms targeting of those destinations but not disruption, and it does not quantify the infected population. Command packets resemble replies from reputable infrastructure, so reputation checks alone can miss malicious traffic. Inventory exposed embedded devices, patch affected vulnerabilities, and remove unnecessary public access. Segment devices that cannot be updated. Use protocol-aware monitoring to investigate abnormal traversal exchanges and unexpected device connections. Check startup configuration and system-utility integrity on suspected devices, and contain confirmed infections as persistent compromises.
Source: https://www.nozominetworks.com/blog/a-stunning-disguise-cling-malware-masquerades-as-google-
10/2/26
LxBase RAT Business Phishing Campaign
MEDIUM-HIGH
+ —
Intel Source:
BI ZONE
Intel Name:
LxBase RAT Business Phishing Campaign
Date of Scan:
10/2/26
Impact:
MEDIUM-HIGH
Summary:
BI ZONE reports mass business-themed phishing campaigns delivering LxBase RAT to Russian companies, with capabilities for credential theft, surveillance, and remote control. Campaigns observed from July through September 2026 targeted finance, engineering, manufacturing, energy, retail, transportation, construction, agriculture, and information technology organizations. Messages impersonated business contacts and used fake orders or quotation requests to lure recipients into opening disguised archives. A multistage script and managed-code chain loads the final payload into a legitimate build process and uses scheduled tasks for recurring execution. The malware can collect browser and email data, authentication material, files, screenshots, and keystrokes. It also supports hidden remote desktop sessions and additional code execution. The report describes delivery and capabilities but does not quantify compromised organizations or confirm use of every supported theft function. Strengthen email attachment inspection and restrict unnecessary execution of scripts received through business correspondence. Use endpoint telemetry to investigate scripts launching build processes, new logon tasks, and unexpected remote-control activity. Contain confirmed infections and assess stolen credentials, session data, and accessible files before restoring normal access.
Source: https://bi-zone.medium.com/lxbase-rat-first-observed-in-attacks-on-russian-companies-a102aec98fa4
10/2/26
Partisan Zmiy Persistence in Healthcare
MEDIUM-HIGH
+ —
Intel Source:
Solar 4RAYS
Intel Name:
Partisan Zmiy Persistence in Healthcare
Date of Scan:
10/2/26
Impact:
MEDIUM-HIGH
Summary:
Solar 4RAYS documents a prolonged Partisan Zmiy intrusion into a medical organization where overlooked security alerts and trusted administrative relationships helped attackers retain access. The investigation began in December 2025 and traced the earliest evidence of compromise to early 2024. Attackers remained for approximately two years without observed destructive activity. The organization had extensive infrastructure and trust relationships with affiliated medical entities. Researchers found a previously undocumented scheduling loader, an updated Vasilek backdoor, service-based persistence, and library replacement within legitimately installed virtualization software. The attackers used messaging services for control, DNS tunnels, and proxy chains, allowing access to continue if one communication channel was lost. The report describes access to sensitive medical data and centralized management systems. Further attacks through affiliated organizations remain a possibility, not a confirmed outcome. Investigate historical security alerts and audit service definitions, virtualization-software integrity, and unusual scheduled execution. Hunt across connected systems for additional backdoors, DNS tunneling, proxy activity, and unauthorized administrative access. Review trust relationships and recovery safeguards, and check for remaining or delayed destructive components before closing the incident.
Source: https://rt-solar.ru/solar-4rays/blog/7027/