A malware campaign targets developers using AI coding assistants by compromising GitHub repositories. Attackers exploit AI agent configuration files like AGENTS.md and CLAUDE.md to deliver a concealed payload (setup.js) that steals credentials from multiple cloud and code platforms. The malware validates stolen tokens, modifies repositories, infects software packages, and propagates by injecting malicious instructions into AI-agent configuration files. It spreads further by committing malicious files through victim accounts with sufficient permissions, impacting major package ecosystems such as npm, PyPI, RubyGems, and JFrog.
CVE-2026-84869 is a critical vulnerability in ConnectWise ScreenConnect that allows unauthorized malicious guests in remote sessions to transfer and execute files on MSP technician workstations with elevated privileges. Attackers exploited this flaw in a large-scale campaign targeting 42 MSP servers, using techniques such as worm-like propagation and path traversal to inject startup scripts and deploy rogue clients. The vulnerability enables full compromise of technician workstations, which hold sensitive RMM consoles and customer credentials. The exploitation campaign included the use of deceptive guest sessions and canary files to verify success before payload deployment.
A phishing campaign targets football fans of EA SPORTS FC 27 by impersonating Prime Gaming reward pages. The fake site captures EA account credentials including email, password, and two-factor authentication codes, sending them to attacker-controlled servers for validation. This enables potential immediate account takeover. The campaign uses advanced credential harvesting techniques and tracks social media traffic sources. Compromised accounts may contain valuable virtual assets, purchased games, and payment information.
Chinese threat actors have been using the Artex agentic AI penetration testing framework to accelerate cyberattacks, including a recent breach of Barracuda. Artex automates complex multi-target operations, enabling attackers with minimal expertise to conduct rapid, large-scale intrusions. The Barracuda attack involved exploitation of exposed cloud access keys, credential theft, lateral movement, and persistence, with attackers accessing sensitive cloud accounts and leaving backdoors in Kubernetes clusters. The framework has been linked to multiple campaigns targeting government, financial, AI development, and cryptocurrency sectors globally. The FBI and security researchers warn that Artex-powered attacks leverage automation to outpace defenders and exploit common security hygiene failures such as exposed credentials and unpatched software.
Between September 30 and October 16, 2026, an AI-driven intrusion campaign targeted AWS environments across four accounts using eight compromised identities. The attacker performed rapid credential validation and harvesting, including 56 EC2 user data requests in 141 milliseconds, extracted secrets from AWS Secrets Manager, and executed remote commands via AWS Systems Manager (SSM) SendCommand. The campaign showed automated behavior with intermittent activity and a deliberate 14-day pause to evade detection. Alerts triggered were individually low or medium severity but became significant when correlated by source IP. The intrusion originated from infrastructure in Hong Kong linked to the open-source CyberStrikeAI offensive security platform.
North Korean state-sponsored hackers have conducted a sophisticated social engineering campaign called FaceStealer, stealing over $400 million in cryptocurrency over 18 months. The attackers compromise legitimate accounts on platforms such as Telegram, LinkedIn, Slack, and Discord to invite targets to fake Zoom or Teams meetings. These calls use stolen webcam footage of previous victims to create the illusion of legitimate participants. Attackers then trick victims into running malicious scripts that install cookie harvesters, keyloggers, screen capturers, and backdoors. The campaign targets cryptocurrency wallet holders, web3 workers, and open-source maintainers, sometimes waiting months or more before stealing funds.
The wfx-loot campaign involved attackers using thousands of hijacked GitHub accounts to test and deploy a credential-stealing malware in over 1,800 repositories. The malware was inserted into GitHub Actions workflows, shell scripts, and npm installation hooks to steal GitHub tokens, repository secrets, cloud credentials, and AI service keys. The attacker used valid tokens rather than exploiting GitHub vulnerabilities and concealed their activity by reversing code changes and deleting workflow runs. Organizations affected should assume all accessible secrets were compromised and take action to revoke and rotate credentials.
The ARTEX attack involves an agentic AI framework that autonomously searches public GitHub repositories for exposed secrets such as OAuth application credentials. Once found, these secrets are abused to access and move laterally within Azure and Microsoft 365 cloud environments, sometimes exfiltrating sensitive data within minutes. ARTEX was publicly available on GitHub until early October 2026 and is now actively used by threat actors in opportunistic campaigns targeting cryptocurrency wallets, gaming accounts, and code-signing certificates. The attack leverages automation and AI to operate at machine speed, significantly accelerating credential abuse and cloud tenant compromise.
Between March and September 2026, the Mirax malware infected thousands of Android devices in the META region, causing millions in estimated financial losses. Distributed via malicious ads in low-reputation streaming apps on Google Play, Mirax impersonates Google Play Services using Cyrillic characters and tricks victims into granting Accessibility Service permissions, enabling full device takeover. It uses server-side configuration with dual-WebSocket command-and-control infrastructure and DNS-over-HTTPS. The malware dynamically pushes phishing templates targeting over 20 regional banks and digital wallets. Attackers remotely control infected devices in real time to perform unauthorized transactions that appear legitimate, with stolen funds transferred to suspected mule accounts in India. The campaign peaked during Eid holidays and remains active.
Ahead of the Grand Theft Auto VI release on November 19, 2026, threat actors are exploiting consumer excitement through phishing campaigns and fraudulent schemes. These campaigns use fake downloads, false early access offers, nonexistent mobile versions, and free game promises to lure victims. Malicious sites distribute trojans, request cryptocurrency payments, use content lockers requiring survey completion, and harvest personal information. Tactics include fake storefronts impersonating Rockstar Games, fraudulent APK files with unsafe permissions, and giveaway scams with urgency counters. Post-launch threats are expected to include in-game currency fraud, credential phishing for Rockstar and gaming platform accounts, and malware through fake cheats and modifications.
The KeyHunter operation is a multi-vector cyber campaign by the Chinese-speaking threat actor leepsmarts targeting academic institutions and cloud infrastructure. It involves theft of large numbers of LLM API keys from misconfigured cloud platforms, creation of fraudulent student accounts at multiple universities, and processing of stolen Harvard alumni data. The operation uses AI agent frameworks and Telegram bots for automation. Parallel financial fraud includes cryptocurrency draining and credit card validation using stolen credentials. The infrastructure spans multiple countries and has been active since early 2026.
MadApe is a sophisticated multi-stage phishing campaign targeting Microsoft Device Code authentication. It uses business-themed PDF lures, configurable redirect chains, and fake document-sharing pages impersonating services like SharePoint, OneDrive, Teams, Adobe Acrobat, and DocuSign. The campaign abuses Microsoft's Device Code flow to access victim accounts without credential theft on phishing pages. It features extensive infrastructure with 155 domains, human-verification challenges, decoy sites, and exposed operator panels. The campaign is localized in at least nine languages and focuses on business communications.
Since July 2025, a financially motivated threat actor named CyberDakait has conducted a sophisticated Android malware campaign targeting India's financial services sector. The campaign uses seven distinct Android trojan families distributed via smishing and vishing, impersonating bank representatives to trick victims into installing malicious APKs. These trojans steal banking credentials, credit card details, and intercept SMS messages including OTPs through overlay attacks. Additionally, the malware performs cryptojacking using XMRig. The malicious apps are hosted on GitHub to appear legitimate, with initial droppers delivering the main payload and mining modules. At least 17,960 devices have been compromised with 7.6 million intercepted SMS messages reported.
A long-running malware campaign has infected 953 online stores, primarily Magento-based, with loader code that selectively serves malicious payloads to real visitors to evade detection. The campaign began in July 2025 and has used multiple domains for its JavaScript loader, with recent activity observed as late as October 2026. The loader performs visitor fingerprinting and has been delivered through exploitation of known Magento vulnerabilities and backdoors. Store owners are advised to check for loader code, vulnerabilities, backdoors, rogue admin accounts, and to apply the latest patches and security measures.
CVE-2026-115191 is a SQL injection vulnerability in the Mautic core product affecting certain versions prior to fixed releases. It occurs in a contact lookup AJAX endpoint of the Mautic administration interface where a request parameter controlling the contact field to search by is not properly validated. Any authenticated user, regardless of role or permissions, can exploit this flaw to execute arbitrary SQL queries against the Mautic database. This allows reading sensitive data such as user accounts, email addresses, password hashes, contact data, and campaign content.
Since September 15, 2026, the Gheist campaign has been abusing stolen GitHub tokens to commit a malicious workflow named sync-check to accessible repositories. This workflow attempts to print repository secrets into logs, which are then exfiltrated via GitHub's API before the logs are deleted. The campaign has affected over 850 compromised accounts and more than 13,500 planted branches, impacting both personal and enterprise private repositories. GitHub's approval controls have prevented exploitation in public repositories, but private and enterprise repositories remain vulnerable. The campaign operates without a command-and-control server, uses the user agent 'gheist', and has multiple evolving variants. Approximately two-thirds of affected accounts have been removed by GitHub, likely as a response measure.
Between July and September 2026, a sophisticated campaign targeted Microsoft 365 accounts using adversary-in-the-middle attacks and token replay techniques. Attackers captured sign-in sessions via AiTM relays and replayed stolen tokens from IP addresses linked to a German hosting provider. The campaign compromised 79 organizations with 497 successful sign-ins amid over 200,000 failed attempts. Malicious actors registered rogue Entra devices, replaced victim MFA methods, and created hidden inbox rules to conceal their activities. The operation evolved through four phases, rotating IP addresses and employing scripted token and browser session replay techniques. Compromised accounts were used for business email compromise, including sending payroll-related emails to external contacts.
Threat actors are conducting phishing campaigns targeting Google Cloud Platform (GCP) environments by using sponsored search ads to steal privileged user credentials. They disable security controls, create service accounts with Gemini API access, and link attacker-controlled external projects to victim billing accounts. This enables unauthorized large-scale AI workload execution, causing significant unexpected charges within 24-48 hours. The attackers use evasion techniques such as geographic IP blending and demoting legitimate administrators to avoid detection and remediation.
Wazza is a multi-stage phishing kit actively targeting Microsoft 365 accounts since July 2026. It uses phishing emails with PDF lures themed around business documents to redirect victims through a complex infrastructure involving Google Sites and CloudFlare Workers. The kit impersonates multiple well-known services and employs advanced browser checks and client-side protections. It abuses Microsoft OAuth device authorization to trick victims into completing authentication, allowing attackers to gain access to accounts and cloud resources. The campaign is medium severity due to credential theft risks.
Incorrect Default Permissions vulnerability in Apache Camel PQC component.
This issue affects Apache Camel: from 4.12.0 before 4.18.5, from 4.19.0 before 4.22.1.
When the component wrote generated key material to disk it did so with the permissions the process default umask produced, rather than restricting the file to its owner. On a host whose umask allows group or world read, the private key was therefore readable by other local accounts. The exposure is to a local party rather than to the remote sender that the Camel threat model treats as the primary adversary, which is why the severity is assessed low, but the consequence where it applies is complete: possession of the private key defeats every signature and key exchange the deployment performs with it.
Upgrade to Apache Camel 4.23.0, which fixes the issue. Users on the 4.18.x LTS releases stream should upgrade to 4.18.5. Users on the 4.22.x LTS releases stream should upgrade to 4.22.1. Key material is now written with owner-only permissions. For deployments that cannot upgrade immediately, place the key directory on a path only the service account can traverse, and tighten the process umask. Key material already written with permissive permissions should be treated as exposed and rotated.