Security news from the BlueTeamSec community for Sunday the 15th of November 2026. A busy one today, with ten stories across threat intelligence, defence tooling, and the continuing evolution of autonomous attack frameworks.
We'll start with threat intelligence. First up, a report from Chinese researchers on APT-C-35, also known as Donot Team or Brain Worm. They've been targeting Pakistani government, military, and diplomatic entities using Eid al-Adha-themed phishing documents that deploy multi-stage malware with encrypted command and control and data exfiltration. Classic intelligence gathering operation with fairly sophisticated evasion techniques.
Staying with North Korean operations, Arctic Wolf has a six-month update on BlueNoroff, the Lazarus subgroup that's now using generative AI to create synthetic personas for social engineering. They've identified 677 targets across 58 countries, overwhelmingly founders and C-suite executives in cryptocurrency and Web3 firms. The industrialisation of the tradecraft is the notable bit here — they're blending malicious activity into routine professional interactions at quite some scale.
And the U.S. Department of Justice has announced disruption actions against two large North Korean theft operations. One called FaceStealer used deepfake-enabled social engineering to steal four hundred million dollars over eighteen months. The other involved Binance-related attacks totalling six billion in theft, with two point seven billion in assets now seized. Both campaigns targeted cryptocurrency exchanges and Web3 individuals to fund weapons programmes.
On to defence tooling. Mihhail Sokolov has released BloodTrail, an open-source performance optimisation for BloodHound CE that replaces the PostgreSQL graph engine with an in-memory one. Complex attack-path queries that took eighty-five seconds on million-node datasets now run in under eight. One for defenders managing large Active Directory estates who need faster threat hunting.
And researchers have developed AgentTrap, a honeypot framework designed specifically to counter autonomous penetration testing agents. It uses stateful, LLM-based deception that adapts to attacker behaviour — in testing, it reduced agent success rates from just under ninety-six percent to seventy-nine percent, and harvested attacker API keys in nearly one in five engagements. Worth a look if you're planning defences against agentic threats.
Moving to exploitation and tooling. The FBI has issued an advisory on ARTEX, an open-source autonomous penetration testing framework that's been weaponised since August. We covered the South Korea bank incident attributed to ARTEX back in early November — this advisory adds detail on a ConnectWise supply chain compromise that affected over two thousand downstream tenants. The FBI notes attackers with minimal expertise can now complete multi-stage intrusions, from reconnaissance to data exfiltration, in under three hours.
Arctic Wolf Labs has observed threat actors using AntiFox, another agentic pentesting platform, to automate attacks against IIS web applications and AWS environments. This follows earlier reporting on similar PHP infrastructure targeting in late October. Attackers deploy web shells, extract credentials from configuration files, then pivot rapidly into AWS API enumeration and secrets exfiltration. The automation significantly compresses the attack timeline.
An academic survey has been published examining the integration of large language models into cybersecurity operations. It covers data requirements, training methods, agent systems, and evaluation frameworks for both defensive and offensive applications. Not incident reporting — this is technical architecture and lifecycle analysis for security-specific LLMs. Useful background if you're assessing where this technology fits in your programme.
Security researchers have released Seep Reverse Lab, an automated reverse engineering workbench that integrates thirty-eight tools to enable AI agents to perform audits across Windows, Android, Linux, and Web platforms. It's focused on identifying client-side authorisation bypass vulnerabilities and comes with nearly three hundred practical notes and defender guidance. GPL-licensed and explicitly intended only for authorised white-box auditing.
And finally, an article from the China Computer Federation discussing forty-five autonomous attack frameworks beyond ARTEX, covering their search targets and discovery methods. It examines how these frameworks operate and techniques for detecting them — adds useful context to the broader landscape we've been tracking since the ARTEX disclosure.
That's everything for today from BlueTeamSec. The original research belongs to the article authors — the analysis here was automated. We'll be back with the next briefing soon.