Forever Security caught an attempt on a customer's machine to subvert an AI coding agent through a planted instructions file. Coding agents read AGENTS.md or CLAUDE.md files in a folder and follow them, and a web page can drop such a file into Downloads without a click. The disguised file told OpenAI Codex to register a server in its global config and install the npm package function-color, part of the MALFEX campaign, which pulls a Windows payload. Against Claude Code, a CLAUDE.md tried to set ANTHROPIC_BASE_URL to an attacker server to leak the login token. Researchers say there is nothing for vendors to patch.
Huntress detailed a Microsoft 365 adversary-in-the-middle campaign that ran from July to September, capturing sign-in sessions behind fake login pages and replaying the stolen tokens from attacker devices within minutes, a technique it calls token spraying. From over 200,000 login attempts the attackers achieved 497 successful sign-ins, compromised 97 accounts across 79 organizations, and registered 48 rogue Entra devices. In at least 19 organizations they removed existing MFA methods and registered their own within three hours, and planted hidden inbox rules hiding payroll and deposit emails. Defenders should revoke sessions, audit Entra devices and MFA methods, and restrict the device-code workflow with Conditional Access.
GitLab disclosed CVE-2026-110264, rated 10.0, which under certain conditions lets an unauthenticated user run arbitrary code on the server through an integer overflow while compiling a specially crafted regular expression. The flaw sits in Onigmo, the Ruby regex engine GitLab ships, and affects every release from 7.0 through 18.11 and the 19.x branches before their fixes, so it spans roughly twelve years of versions. Only self-managed installations must act; GitLab.com and Dedicated already run patched builds. Fixes are in 18.11.13, 19.0.10, 19.1.9, 19.2.9, 19.3.6, 19.4.3, and 19.5.1. GitLab reports no exploitation yet.
Researchers at SafeDep, Socket, StepSecurity, and Upwind reported that LiteLLM version 1.107.0, a popular Python AI gateway, was published with malware through compromised GitHub CLI OAuth tokens belonging to three engineers. A cook.pth file runs a payload whenever the Python interpreter starts, even before LiteLLM is imported. The first stage, calling itself privcollect, harvests cloud tokens, developer credentials, private keys, AI tooling credential files, browser data, environment variables, and the proxy's saved keys and model config, exfiltrating over HTTP to litellm[.]dev/healthcheck. A second payload is a backdoor that fetches and runs further commands. The PyPI package and Docker images were pulled, and users should pin to 1.106.2.
A researcher found more than 300 malicious Tor exit relays, at their October 4 peak about 7.46 percent of exit capacity, rerouting SSH connections to twelve interception servers. The attackers edited relay torrc files to allow port 22, and each hijacked exit presented its own host key before connecting onward to the real server, so sessions worked while passwords typed by users who accepted the key were captured. Key-based logins could not be relayed. The operator of the largest affected relay family traced the root cause to the Ebury backdoor and a reused plaintext password with SSH password authentication left enabled. Tor flagged the relays by October 5.
TeamUpdraft disclosed that attackers breached its website over the October 17 weekend and reached its customer database, obtaining names, emails, hashed passwords, 2FA settings, and the UpdraftCentral connection keys that let them push backdoor plugins to connected WordPress sites. UpdraftPlus, a backup plugin with over four million installs, ships an UpdraftCentral client in every copy. Patchstack saw signed plugin.upload_plugin commands over an 18-hour window and 99 backdoor plugin names, and one backdoor creates a persistent data-ops43 editor account that survives plugin removal. Fixes are in UpdraftPlus 1.26.13 and 2.26.13, which disconnect pre-incident UpdraftCentral connections. Site owners should rotate keys and hunt for rogue accounts.
PHP shipped emergency patches on October 21 for two runtime flaws collectively called PHP2Shell, now confirmed as CVE-2026-109618 and CVE-2026-109619 and rated 9.2, correcting the leaked 10.0 figure. They affect nearly every PHP version back to 5.4, with fixes in 8.5.12, 8.4.27, 8.3.36, and 8.2.35; branches 8.1 and older get no official update. Both live in the PHP runtime, so applications are exposed even with secure code, and the risk is highest under Apache mod_php. Cloudflare and watchTowr reported exploit attempts within hours, though no confirmed compromise. Wiz, Calif, and OpenAI's Codex Security team reported the flaws.
CISA added six actively exploited flaws to its Known Exploited Vulnerabilities catalog. The newest is CVE-2026-106931, a command injection rated 8.5 in EG Secure Solutions SiteGuard Server Edition, a Japanese product the vendor confirmed is being exploited. CISA also added the four Plex2Shell flaws in Plex Media Server, which researchers chain for unauthenticated remote code execution and which are being used to deploy cryptocurrency miners and proxy malware, and the Siemens Software Center DLL sideloading flaw CVE-2026-110936 abused in ClickFix attacks. Federal agencies must patch SiteGuard and two Plex flaws by October 23, and the remaining ones by November 3.
Mozilla and Google's Chrome Root Program confirmed that the private keys of eight eMudhra intermediate certificate authorities were compromised, six TLS intermediates and two code-signing, after a researcher reported it on October 17 and proved control by signing messages with each key. A key holder could issue a certificate for any website a browser would accept, though Chrome found no affected TLS certificate logged to Certificate Transparency and reported no confirmed misissuance. Chrome, Mozilla, Apple, and Microsoft blocked the intermediates within days, and Go shipped releases 1.27.3 and 1.26.10 under CVE-2026-111943 that reject any chain containing the keys. Firefox and other projects issued fixes.
Fortinet patched CVE-2026-106173, a critical path traversal flaw rated 9.4 in FortiSIEM that lets an unauthenticated attacker execute commands through specially crafted packets, and researchers have now released a public proof of concept. The attacker connects to the gRPC service on TCP port 27918 and writes a file outside the intended directory, using a null byte in the query-id parameter to truncate the filename so the write lands on an existing path such as a Redis health-check script that cron runs every 60 seconds. FortiSIEM 7.5.0 is affected, 7.6 is not, and users should upgrade to 7.5.1 or above. Horizon3.ai published the research.