Loki Intelligence — Security Briefs · Published
Daily Security Brief: Linux Kernel Root Risks, AI Agent Plugins, and Cisco ISE
Linux kernel privilege escalation dominates the day, with three exploited kernel bugs added to KEV and public exploit code released for four others, while AI tooling shows new supply-chain and cloud privilege risks. Teams should accelerate kernel patching through distribution channels, update AI coding agents and review their plugin sources, and finish Cisco ISE remediation.
Signal 01 · The Hacker News
Microsoft Patches CVSS 10.0 Azure AI Foundry Flaw Enabling Unauthorized Privilege Escalation
Microsoft fixed CVE-2026-85889, a CVSS 10.0 missing-authentication flaw in Azure AI Foundry that could allow network-based privilege escalation, along with critical issues in Microsoft 365 Copilot, Azure Database for PostgreSQL, and Cosmos DB. The cloud flaws are already mitigated, and Microsoft reports no exploitation of the Foundry bug.
Why it matters: No action is needed for the cloud-side fixes, but apply the out-of-band Windows 11 26H1 update that addresses two local privilege escalation flaws.
Source: The Hacker News
Signal 02 · CISA
CISA Adds One Known Exploited Vulnerability to Catalog
CISA added CVE-2025-39682, a Linux kernel flaw involving an improper check for unusual or exceptional conditions, to the KEV catalog on evidence of active exploitation. The catalog entry ties it to the kernel's TLS receive path, where zero-length records can break expected record handling.
Why it matters: Track your distribution's kernel advisories for this CVE, prioritize internet-facing and multi-tenant Linux hosts, and confirm machines rebooted into patched kernels.
Source: CISA
Signal 03 · CISA
CISA Adds Two Known Exploited Vulnerabilities to Catalog
CISA also listed two more exploited Linux kernel bugs: CVE-2025-39964, a race condition in AF_ALG sockets, and CVE-2026-53266, an out-of-bounds write in the ebtables SNAT target. CISA notes some affected products may be end-of-life, and federal agencies have until September 21 to act.
Why it matters: Inventory kernel versions across servers, appliances, and container hosts, retire unsupported kernels, and patch on an accelerated schedule given three kernel KEV additions in one day.
Source: CISA
Signal 04 · The Hacker News
Cisco Warns of New Zero-Day ISE Auth Bypass (CVSS 10.0) Exploited in Active Attacks
Cisco's maximum-severity ISE authentication bypass, CVE-2026-76460, remains under active exploitation and is now on CISA's KEV list. Successful attacks may yield root command execution, and because attackers could erase evidence, Cisco recommends re-imaging affected nodes rather than attempting simple cleanup.
Why it matters: Confirm every ISE and ISE-PIC node runs a fixed patch, check logs across distributed deployments, and plan rebuilds wherever compromise indicators appear.
Source: The Hacker News
Signal 05 · The Hacker News
Plugin4Shell Lets Repository Owners Swap Pinned Plugin Code Across Four AI Coding Agents
Air Security found four AI coding agents installed plugins pinned by commit hash without verifying the fetched code matched, letting a repository owner swap in different code on hosts that allow hash-like branch names. Air says Claude Code 2.1.179 and Codex 0.146.0 fix it, Copilot has no fix, and Gemini CLI will not be patched.
Why it matters: Update coding agents, install plugins only from trusted GitHub-hosted marketplaces, avoid auto-update for third-party sources, and remember plugins run with the developer's full access.
Source: The Hacker News
Signal 06 · The Hacker News
CISO's Expert Guide to Agentic Pentesting for Websites
Reflectiz's sponsored CISO guide makes the case that continuous, agent-driven web pentesting can close the gap between days-long attacker timelines and weeks-long patch cycles. It urges buyers to demand enforced test coverage, separate validation of findings, and safety controls before letting an autonomous agent test production.
Why it matters: Evaluate agentic testing vendors on verifiable coverage and safe-stop controls, and keep human approval over scope, especially when testing business-logic flaws such as IDOR.
Source: The Hacker News
Signal 07 · The Hacker News
Public Exploits Released for Four Linux Kernel Flaws That Enable Local Root
Researcher Asim Manizada published working root exploits for four Linux kernel networking flaws, DirtyAH6, TUNderflow, PPPoEject, and DiagSpill, after coordinated fixes landed in stable kernels. No in-the-wild use is reported; three need unprivileged user namespaces, while DiagSpill only requires the SCTP module to be available.
Why it matters: Install distribution kernels carrying all four fixes; until then, consider disabling unprivileged user namespaces and unused AH6, TUN/TAP, PPPoE, and SCTP features on multi-user hosts.
Source: The Hacker News
Brief sources
Related briefs
- What Is AI Agent Security Testing?
- AI Red Teaming vs Traditional Pentesting
- MCP Security Testing Checklist
Relevant Loki service: Orvyn — AI agent security private preview.