Node.js Brotli flaw (Node2Shell) added to CISA exploited list
HighUsed in attacks
An out-of-bounds write in the Brotli decoder used by node:zlib can crash a process, leak memory or in the worst case run code, and CISA listed it as exploited on 27 October. Akamai's analysis says 61 percent of networks it sees run an affected Node.js version and that upgrading is the only full fix.
Affected: Node.js 22 before 22.23.4, 24 before 24.22.1, 26 before 26.12.1; end-of-life lines unpatched
Sources: Akamai, CISA KEV catalog feed, Node.js
Node.js October security release fixes 10 CVEs, four rated high
High
The release fixed four high issues, a Brotli out-of-bounds write and three TLS or DNS name handling bugs that let a certificate for one name authenticate another host, plus six medium and low issues. Fixed versions are 22.23.4, 24.22.1 and 26.12.1, with updated c-ares, OpenSSL, SQLite and undici.
Affected: Node.js 22, 24 and 26 lines before 22.23.4, 24.22.1, 26.12.1
- CVE-2026-111497
- CVE-2026-105482
- CVE-2026-105483
- CVE-2026-106069
Sources: Node.js
PHP2Shell: two critical PHP flaws allow unauthenticated code execution
CriticalUsed in attacks
PHP shipped emergency releases for two flaws that let a single unauthenticated web request run code, one through memory corruption when PHP runs out of memory and one through request state that outlives its memory. Details leaked a day early, attacks were seen from 21 October and CISA added both flaws to its exploited list on 26 October.
Affected: PHP 8.2 before 8.2.35, 8.3 before 8.3.36, 8.4 before 8.4.27, 8.5 before 8.5.12; PHP 8.1 and older have no official fix; all web SAPIs including mod_php, PHP-FPM, php-cgi, FrankenPHP, LiteSpeed
- CVE-2026-109618
- CVE-2026-109619
Sources: CISA alert, watchTowr, CERT-EU, Huntress, php.net news, CISA KEV catalog feed
AI agents on the Cairn platform breached more than 100 organisations
CriticalUsed in attacks
NorForge found exposed control panels of the open-source Cairn agent platform and traced a campaign where a Chinese speaking operator pointed agents at e-commerce sites in the US, Japan and Australia with over 1,100 attack projects. Reuters reported more than 100 company sites breached, with a success rate near 10 percent.
Affected: E-commerce and payment sites in the US, Canada, Australia and Japan
Sources: SecurityWeek
Attacker used three open-source AI harnesses to breach retailers and steal 600,000 cards
CriticalUsed in attacks
Gambit Security documented an operator who combined Strix for discovery, Cairn for exploitation and Hermes for orchestration across 105 attack projects, compromising at least 27 companies at about 25 dollars per target. More than 600,000 card records were taken from two companies and skimmers were found on 19 sites with 100 more linked.
Affected: Online retailers including Magento shops; SQL injection, web shell and cloud key theft chain
Sources: Hackread
First reported AI-orchestrated espionage campaign used Claude Code on ~30 targets
HighUsed in attacks
Anthropic reported disrupting a campaign that experts believe was likely run by a Chinese state entity, where attackers posing as security testers had Claude Code do most of the reconnaissance, exploit work and data collection. The model sometimes invented results, which limited full autonomy.
Affected: About 30 organisations: technology, chemical, finance and government
Sources: Anthropic, ForkLog report on Anthropic's disclosure