Daily · 2026-05-08
🐧 Dirty Frag Linux Root, 🎓 Canvas Defaced, 🤖 ClaudeBleed Takeover, 🔥 PAN-OS KEV, 📦 Trellix RansomHouse, 🐛 Next.js Vuln Cluster
Another universal Linux root zero-day (Dirty Frag) hits with no patches, ShinyHunters defaces 330 Canvas login portals in second Instructure breach, and AI agent security cracks widen with ClaudeBleed Chrome extension takeover and Semantic Kernel prompt-injection RCE — prioritize Linux kernel module mitigations, Canvas breach notifications, and AI agent framework patching.
Dirty Frag is the day's defining story. Dirty Frag is a deterministic Linux kernel logic bug in the Dirty Pipe and Copy Fail family: it chains two page-cache write flaws to give local users root on all major distributions with a single command, and the proof of concept was public before any patch. A universal root exploit with no fix is a race against every Linux fleet.
The second Instructure breach is the most visible escalation. ShinyHunters defaced roughly 330 Canvas login portals in a new intrusion, demonstrating retained access despite prior patching, and set a May 12 extortion deadline tied to the 275 million-record claim. Affected schools should assume data exposure and prepare for direct extortion contact.
The AI-agent surface produced three findings in one day: ClaudeBleed lets any installed extension drive the Claude extension through origin-based trust, Semantic Kernel prompt injection reaches host-level code execution, and Claude Code OAuth tokens are stealable through MCP hijacking via npm lifecycle hooks. Ivanti's EPMM zero-day is under active exploitation with over 850 exposed instances, and CISA added PAN-OS CVE-2026-0300 to KEV with hallmarks pointing at Chinese state-sponsored activity.
Watchlist: Dirty Frag patches as they land, Canvas breach notifications, AI agent extension and MCP trust boundaries, and EPMM patching.