11 stories · 5 sources · 1 critical · 2 high · ~10 min read

Coverage: Last 24 hours

Today’s Highlights

Credential compromise, critical Linux kernel flaws, and live AI-enabled threats highlight major operational risks for defenders. Patch status, real-world exploitability, and hidden persistence are central defender concerns throughout this cycle. The most urgent stories include the compromise of the Claude Code VS Code extension in a supply chain attack and Google’s Gemini 4 Carbon taking a lead in automated security benchmarks. Attackers exploit trusted AI platforms and developer toolchains, underscoring the need for continuous monitoring and aggressive response.

Table of Contents

1. Linux Kernel TCP Flaw With 9.8 CVSS Score Affects Ubuntu, Red Hat and Other Distributions
2. Official Claude Code VS Code Extension Compromised to Install Persistent SSH Backdoor
3. ThreatsDay: Critical Firewall Flaws, Banking Trojans Rented, Weaponized PDF, Token Theft + 26 More Stories
4. Google’s Gemini 4 Carbon Leads Benchmarks for Finding, Exploiting, and Patching Vulnerabilities

Critical   High   Medium   Low

Emerging Signals

Linux Kernel TCP Flaw With 9.8 CVSS Score Affects Ubuntu, Red Hat and Other Distributions

Source: The Hacker News | Published: Oct 29 | Risk: CRITICAL | Impacted: Linux servers with multipath TCP enabled, Internet-facing Ubuntu and Red Hat deployments, Managed hosting environments | Topics: Vulnerability / Exploit

What happened: A critical flaw in the Linux kernel’s TCP code, identified as CVE-2026-98432, has been discovered, affecting distributions like Ubuntu and Red Hat Enterprise Linux. This vulnerability, rated 9.8 out of 10 on the CVSS scale, allows remote attackers to execute commands as root on servers with multipath TCP listeners. As of October 29, 2026, Ubuntu and Red Hat have not released patches for this issue.

Why it matters: Unpatched, internet-facing Linux systems, especially those using multipath TCP, are exposed to unauthenticated remote code execution as root, enabling full system compromise at scale.

How it works: This vulnerability resides in the multipath TCP implementation in the Linux kernel. An attacker can craft TCP packets to exploit the flaw, leading to remote code execution as root without the need for an existing account or local presence.

Affected / Fix: Affects Ubuntu and Red Hat; no official patches available as of October 29, 2026.

Practitioner Perspective

Server teams running any affected Linux distribution need to urgently assess exposure to CVE-2026-98432. This is a classic kernel-level RCE where remote, unauthenticated attackers can completely take over systems without user interaction. It poses operational risk beyond direct exploitation, as a foothold on even one critical host can enable lateral movement across data centers and clouds. Where patches are not yet available, compensating controls like network segmentation and strict ingress filtering are mandatory. Do not delay mitigation while waiting for vendor patches: prioritize defense-in-depth and out-of-band containment steps.

Recommended Actions

• Scan all Linux systems for multipath TCP listener exposure to CVE-2026-98432
• Apply available mitigations or disable multipath TCP on Ubuntu and Red Hat servers until patches are issued
• Segment affected hosts from the general network and restrict incoming connections
• Aggressively monitor for suspicious kernel-level or root activity linked to network sources
• Initiate threat hunting for exploitation attempts using public PoCs against this kernel flaw

Official Claude Code VS Code Extension Compromised to Install Persistent SSH Backdoor

Source: The Hacker News | Published: Oct 30 | Risk: HIGH | Impacted: Developer endpoints running VS Code, Organizations using anthropic.claude-code, Teams with unmonitored SSH access, Software development and DevOps environments | Topics: Vulnerability / Exploit

What happened: The official Claude Code VS Code extension was compromised to install a persistent SSH backdoor on infected hosts. The malicious version, identified as anthropic.claude-code version 2.1.313, added an attacker-controlled SSH key labeled ‘svc-telemetry’ to users’ ‘~/.ssh/authorized_keys’ files. This allowed the attacker to establish a reverse SSH tunnel to ‘134.209.82[.]225’ and transmit host reconnaissance data over an unencrypted HTTP request. The malicious extension was removed, and version 2.1.314 was published. Users who installed the compromised version between 2:27 a.m. and 1:35 p.m. PT on October 29, 2026, are advised to isolate their hosts, remove the unauthorized SSH key, and rotate credentials and sessions.

Why it matters: Developer workstations may already be persistently compromised, giving attackers ongoing SSH access for lateral movement and data exfiltration, especially in organizations that rely on VS Code and its extensions without strong endpoint controls.

How it works: The Claude Code VS Code extension enhances Anthropic AI capabilities within the VS Code IDE. The compromised version silently adds an attacker-controlled SSH key, allowing remote shell access and data exfiltration, effectively bypassing most network defenses.

Affected / Fix: Malicious code was removed in version 2.1.314; users must upgrade if they installed 2.1.313 during the Oct 29, 2026 compromise window.

Practitioner Perspective

Any team that had developers or admins install anthropic.claude-code v2.1.313 in the exposure window should treat those hosts as potentially backdoored. This attack leverages a trusted supply chain node (VS Code extensions), allowing attackers to bypass traditional phishing or malicious download prevention. With SSH credentials and persistence established, attackers can easily escalate privileges, pivot to sensitive internal environments, or exfiltrate intellectual property. Begin incident response workflows, focusing on credential rotation and SSH trust relationships. The highest priority should be remediating affected developer endpoints and verifying all sources of extension updates.

Recommended Actions

• Search all developer systems for the ‘svc-telemetry’ SSH key in ~/.ssh/authorized_keys
• Hunt for outbound SSH connections and HTTP traffic to 134.209.82.225 from affected hosts
• Immediately upgrade anthropic.claude-code to version 2.1.314 or later and remove 2.1.313
• Rotate SSH keys and credentials on any system where the compromised extension was present
• Verify recent VS Code extension installations for unexpected publisher or update sources

ThreatsDay: Critical Firewall Flaws, Banking Trojans Rented, Weaponized PDF, Token Theft + 26 More Stories

Source: The Hacker News | Published: Oct 29 | Risk: HIGH | Impacted: Organizations running legacy firewalls, Teams with lax PDF and SaaS file security, Cloud environments relying on token-based auth | Topics: Malware / Vulnerability

What happened: The Hacker News reports on various cybersecurity incidents, including critical firewall vulnerabilities, rented banking trojans, weaponized PDFs, token thefts, and more.

Why it matters: A surge in serious vulnerabilities and malware across multiple platforms increases the likelihood of exploitation across poorly maintained systems where patch cycles or compensating controls lag behind threat actor tactics.

How it works: The report highlights a variety of active exploits: perimeter device vulnerabilities, PDF-based malware delivery, cloud credential theft, and malware-as-a-service targeting financial systems. Weak controls or delayed patching can allow attacker chaining and persistent access.

Affected / Fix: Multiple product advisories released; details in referenced report per platform.

Practitioner Perspective

The clustering of critical firewall flaws, commodity banking trojan campaigns, and token-focused attacks reflects active targeting of misconfigured or outdated perimeter and cloud services. Defenders should avoid a siloed response: attackers can chain these weaknesses for broader compromise. Response teams must tighten vulnerability management SLAs, prioritize CVEs reliably under weaponization, and track cross-platform malware propagation closely. The most important consideration is linking indicators across technology silos to preempt lateral movement.

Recommended Actions

• Inventory and urgently patch firewall appliances cited in the referenced critical vulnerability list
• Update detection rules for weaponized PDF malware strains active in the reporting period
• Hunt for bank trojan IOCs, especially in endpoint logs from at-risk business units
• Review token management audit logs for suspicious activity associated with external SaaS integrations

Exploits & CVEs

Google’s Gemini 4 Carbon Leads Benchmarks for Finding, Exploiting, and Patching Vulnerabilities

Source: The Hacker News | Published: Oct 29 | Risk: MEDIUM | Impacted: Organizations automating security testing with LLMs, Development teams integrating AI-powered code review, IR and vulnerability management leads | Topics: Vulnerability / Ai

What happened: Google on Wednesday took the wraps off Gemini 4 Carbon, its latest frontier model with state-of-the-art cybersecurity and long-horizon capabilities. “It shows significant gains across coding and agentic tasks as well as across key capabilities critical for cybersecurity defense, landing a firm lead in SEC-bench Pro (88.2%) for finding vulnerabilities, ExploitGym (54.8%) for exploitation, and”

Why it matters: The emergence of cybersecurity-specialized AI models with state-of-the-art vulnerability finding and exploitation capabilities poses both a defensive opportunity and an automation risk if not carefully governed.

How it works: Gemini 4 Carbon is an advanced AI model designed for cybersecurity tasks, including code analysis, vulnerability discovery, and even automated exploitation and remediation. These capabilities can accelerate both defensive and offensive operations.

Affected / Fix: Gemini 4 Carbon publicly announced as of October 29, 2026; organizations must vet adoption strategy.

Practitioner Perspective

Google’s Gemini 4 Carbon demonstrates that specialized AI is now effective at not only flagging, but also exploiting and patching vulnerabilities in code at scale. Security teams must evaluate the defensive utility of such models, but also anticipate how attackers could leverage similar automation for rapid exploit development. Carefully consider how and where to deploy AI-enabled security controls, and monitor for changes in adversary tradecraft linked to automated exploit generation. The most critical need is for SOC and IR teams to validate model-driven outputs before taking action.

Recommended Actions

• Evaluate Gemini 4 Carbon’s vulnerability finding and exploitation benchmarks for internal adoption
• Pilot AI-driven code analysis in controlled environments before expanding to production codebases
• Monitor for adversary use of AI-generated exploits, especially in vulnerability research and bug bounty communities
• Integrate human review checkpoints where Gemini 4 Carbon or similar models inform patch or mitigation decisions

Also Today

• Fake ChatGPT and Claude Apps Abuse Signed Executables to Deploy Crypto Miners: Threat actors deliver malware and cryptocurrency miners disguised as AI desktop apps for ChatGPT and Claude across GitHub.
• More than half of Victorian lawyers use AI – but only a quarter have cut client fees as a result: A majority of Victorian lawyers are using AI, but only a minority have reduced client fees; AI use increases error correction workload.
• What is it like to work when your co-workers are AI agents?: Companies are delegating operational roles to AI agents, raising privacy and reliability concerns among employees.
• Could Abdul El-Sayed’s plan for AI companies stave off catastrophe? | Jerel Ezell: Senatorial candidate Abdul El-Sayed calls for public ownership and reforms in the AI sector to reduce systemic risk.
• The Download: AI agent tracking and climate feedback loops: Experts urge for real-time monitoring of AI agents’ actions to improve governance and reduce unintended consequences.
• Mistral Says Its AI Model ‘Le Chonk’ Has Leapfrogged China in the Open-Weight Race: The Mistral ‘Le Chonk’ AI model claims to outperform leading Chinese models in coding and security benchmarks.
• An updated Model Spec for more capable and beneficial AI: OpenAI releases a new Model Spec emphasizing human intent, transparency, and harm minimization in AI model behavior.

Defensive Actions

• Search all developer systems for the ‘svc-telemetry’ SSH key in ~/.ssh/authorized_keys to detect the compromised Claude extension.
• Hunt for outbound SSH connections and HTTP traffic to 134.209.82.225 from potentially affected developer endpoints.
• Immediately upgrade anthropic.claude-code to version 2.1.314 or later and remove all installations of version 2.1.313.
• Scan all Linux systems for multipath TCP listener exposure to CVE-2026-98432, and apply mitigations or disable multipath TCP until vendor patches are available.
• Inventory and urgently patch firewall appliances cited in the latest vulnerability lists; update rules for active malware strains seen in this cycle.
• Evaluate Gemini 4 Carbon’s performance in internal AI-augmented security workflows, and integrate human review checkpoints where model outputs drive mitigation or patch actions.
• Block installation and execution of GPT-5.6 Sol, GPT-6 Astra, and Claude Opus fake desktop applications, and audit endpoints for anomalous CPU or cryptominer signatures.

What We’re Watching

• Continued exploitation of Linux kernel CVE-2026-98432 (CVSS 9.8) across unpatched distributions; expect weaponized public exploits to emerge rapidly.
• Ongoing risk from VS Code supply chain compromises, especially targeted at developer endpoints through extension updates.
• Advancements in adversarial and defender use of AI models like Gemini 4 Carbon and Mistral Le Chonk for vulnerability research and automated exploitation.
• Expansion of banking trojan and token theft campaigns leveraging new delivery vectors including PDFs and SaaS integrations.
• Regulatory developments or major vendor responses to emergent issues in AI agent transparency and model safety specifications.

Found this briefing useful? Follow the blog to get the next one as soon as it is published, and pass it along to a colleague who owns patching.

Categories: Artificial Intelligence, Cybersecurity Blog

Tags: AI Security, Exploit, Linux, Malware, Supply Chain