backdoor

Research · since 13 Nov 2026

Researchers introduce a stealthy finetuning-activated backdoor attack that implants a dormant backdoor into pretrained vision encoders. The attack bypasses limitations of existing methods under full-model finetuning.

Breaches & attacks · Exploited · since 12 Nov 2026

Multiple threat actors are actively abusing the legitimate GSocket tool to maintain persistent access on compromised Linux servers. The technique allows attackers to retain control even after underlying initial access vulnerabilities are patched.

Do:Audit Linux servers for unauthorized GSocket installations and inspect network connections.

Research · since 11 Nov 2026 · updated 13 Nov 2026

Microsoft researchers demonstrated that AI models can be backdoored by editing fewer than 10 rarely used neurons without retraining. This triggers malicious commands that connected AI agents can execute with their permissions.

Research · since 10 Nov 2026

Researchers demonstrated a task-level backdoor attack against self-evolving LLM agents by implanting malicious rules through failure-aware task optimization. The attack compromises persistent skills without requiring direct manipulation of victim components.

Breaches & attacks · Exploited · since Oct 2026 (approx.) · CVE-2026-105134, CVE-2026-105133

Researchers observed attackers exploiting CVE-2026-105133 and CVE-2026-105134 in AhsayCBS to install a Python-based backdoor. The chain bypasses authentication and runs arbitrary commands on compromised systems.

Do:Apply the released patches for CVE-2026-105133 and CVE-2026-105134 and monitor for the backdoor

Research · since 9 Nov 2026

Researchers demonstrated reference-feedback poisoning in text-space skill optimization, showing that malicious reference answers can induce persistent trigger-conditioned rules in agent skills. The attack succeeds without requiring direct access to the model weights or optimizer.

Supply chain · Exploited · —No fix needed · since 6 Nov 2026

Threat actors are hosting malicious security projects on GitHub to target cybersecurity professionals. The fake repositories deliver infostealers and backdoors to unsuspecting users.

Do:Avoid downloading unverified security tools from untrusted GitHub accounts.

Supply chain · Exploited · —No fix needed · since 5 Nov 2026 · updated 6 Nov 2026

Threat actors are distributing a new Linux backdoor named CWPdoor via a trojanized version of the Control Web Panel web server. The malicious binary enables attackers to execute commands silently without generating logs.

Do:Check servers for unauthorized modifications to the cwpsrv binary and remove any instances of CWPdoor.

Supply chain · Real loss · ●Patched · since 2 Nov 2026 · updated 4 Nov 2026

Attackers breached the update system of the cPGuard security tool on November 2 to distribute a malicious script to customer Linux servers. The script deployed a backdoor and exfiltrated sensitive files including password hashes.

Do:Update cPGuard to version 5.88.25 or later and inspect servers for compromise.

Breaches & attacks · Exploited · since 30 Oct 2026 · updated 4 Nov 2026

Google successfully disrupted a 1.3 million-device Android botnet operating primarily on TV set-top boxes. Investigators attribute the botnet to device maker Videostrong, which allegedly integrated backdoors to sell unauthorized network access.

Breaches & attacks · Real loss · since 2 Nov 2026

Google dismantled the infrastructure of the XCPE botnet, which infected over 1.3 million Android TV set-top boxes via embedded firmware backdoors. The hardware manufacturer allegedly sold unauthorized access to criminal affiliates.

Breaches & attacks · Real loss · since 25 Oct 2026 · updated 3 Nov 2026

Google and industry partners dismantled the XCPE backdoor which infected over 1.3 million set-top boxes. The OEM Videostrong Technology operated the botnet for financial gain by loading ad-fraud modules.

Do:Disconnect or replace affected compromised set-top boxes.

Supply chain · Exploited · ○No patch yet · since 30 Oct 2026 · updated 2 Nov 2026

A tampered build of Anthropic’s Claude Code extension planted an SSH backdoor on developers’ machines via the Open VSX marketplace. Researchers discovered the malicious version 2.1.313 distributing unauthorized access.

Do:Remove the malicious Claude Code extension version and check systems for unauthorized SSH keys.

Research · since 30 Oct 2026

Research · since 21 Oct 2026

Research · since 21 Oct 2026

Research · since 20 Oct 2026

Research · since 19 Oct 2026

Research · since Jun 2025 (approx.)

Research · since 18 Mar 2025

Research · since 31 Mar 2023

Research · since 18 Jan 2021