RC RANDOM CHAOS

PHP Patches Two Critical RCE Vulnerabilities After Months of Delays

· via Hacker News

Original source

Today's PHP security release, and how we're strengthening security

Hacker News →

PHP released critical patches for versions 8.5.12, 8.4.27, 8.3.36, and 8.2.35 addressing two remote code execution vulnerabilities. CVE-2026-109618 stems from memory corruption when a request hits PHP’s memory_limit at precisely the wrong moment, leaving partially initialized data treated as valid—exploitable before application code runs on Apache with mod_php. CVE-2026-109619 occurs when a fatal error leaves internal state pointing to freed memory, allowing later requests on the same worker to trigger use-after-free conditions. Both vulnerabilities sit in core systems—memory management and error handling—that nearly every request exercises, making fixes difficult without breaking compatibility or requiring third-party extension changes.

A PHP core developer contracted by the PHP Foundation discovered the OOM problem in 2024 while fixing issues from that year’s external security audit by Quarkslab, funded by the Sovereign Tech Agency. Although a theoretical exploit was theorized that year, concrete attack vectors remained unknown, and the security team faced multiple conflicting approaches—each with trade-offs in compatibility, denial-of-service risk, or requiring third-party extension modifications. The small team with deep engine knowledge struggled to choose between approaches while managing competing responsibilities, causing work to stall and restart repeatedly. The actual severity emerged in August 2026 when Calif reported an unconditional RCE on the 11th, immediately followed by OpenAI’s independent discovery and further attack vectors from Wiz on August 24th and OpenAI on September 18th.

The PHP Foundation restructured its security operations, creating an Ecosystem Security Team in May 2026, bringing Volker Dusch on as AI Security Engineer in Residence, and recruiting Alexandre Daubois in July and Daniel Scherzer in September for triage and hardening work. The foundation enrolled PHP in the Patch the Planet program and expanded security tools beyond an earlier Anthropic Claude Mythos scan to include OpenAI’s Daybreak Red. Since July 1st, the team published 15 CVEs while processing over 100 invalid, duplicate, or irreproducible reports—a volume increase driven by AI-assisted research. Coordinated disclosure fractured when CVE IDs leaked on October 20th at 15:00 UTC with incorrect details, prompting an accelerated pre-announcement at 19:30 UTC and a release advance to October 21st. Calif’s mistakenly published proof-of-concept video and Chainguard’s unilateral embargo break at 07:41 UTC on release day further complicated the disclosure.

Read the full article

Continue reading at Hacker News →

This is an AI-generated summary. Read the original for the full story.