Bring Your Own EDR: Turning a Commercial Endpoint Agent Into a Trojan Horse

Akamai researchers detailed how a legitimately deployed commercial EDR can be flipped into an attacker’s implant, the standout offensive research of the day. Elsewhere, fresh PoCs and loaders shipped, Clop’s Windchill campaign widened to Shell and Philips, and the RingCentral breach spilled 1.6M records onto Have I Been Pwned.

Offensive & Red Team

• “Bring Your Own EDR” — Akamai’s security research team walked through how a trusted, commercially deployed endpoint detection agent can be weaponized as a covert delivery and persistence mechanism — a trojan horse that rides in on software defenders already trust and whitelist. The technique inverts the usual EDR-evasion problem: rather than dodging the sensor, the operator co-opts it. Akamai

New Tools & Releases

• Dark Agent COFF loader — BadSector Labs published a macOS/Linux implant with a COFF loader, a comparatively rare drop for offensive tooling outside the Windows ecosystem. Useful for red teams building cross-platform post-exploitation capability. minorimpact.dev
• Third Linux privilege-escalation PoC from aramosf — the researcher released another local privesc proof-of-concept, again derived from reverse-engineering the commit that silently fixed the bug — a reminder that upstream kernel patch diffs remain a reliable exploit-primitive source. @aramosf via @Dinosn
• Gajim single-click RCE PoC (CVE-2025-49091) — public exploit targeting the Gajim XMPP client on KDE Plasma systems, triggered by a single click. GitHub PoC

Vulnerabilities & Exploits

• Adobe Commerce / Magento CVE-2026-71362 now has a reproducible lab — a Docker environment reproducing the customer-session identity-switch account-takeover flaw (APSB26-92, CVSS 9.1) was published, lowering the bar to study a bug already seen exploited within hours of disclosure (earlier coverage). @Dinosn

Threat Activity

• Clop’s Windchill campaign balloons to 40+ named victims — the gang listed dozens of new companies on its leak site, including Shell, Philips, and GE, consistent with mass exploitation of PTC Windchill (earlier coverage). @campuscodi (discussion)
• Lazarus hid a zero-day behind post-quantum cryptography — Denmark’s CERT.dk reported that the DPRK group concealed exploitation of a 0-day flaw using post-quantum crypto to frustrate analysis and detection. CERT.dk
• KB-Backdoor — a ~12KB custom Windows backdoor posing as a Realtek audio component was found persisting via a daily WMI scheduled event and hiding its C2 domain inside whitespace characters in a desktop.ini file. It beaconed over ICMP and HTTP and stayed active long after its C2 domain expired — a deliberately low-profile design that defeats naive static and network detection. @blackorbird
• Mustang Panda’s CoolClient goes kernel-mode — researchers discovered a rootkit variant of the CoolClient implant, using a kernel-mode driver, in victims across Pakistan, Mongolia, Myanmar, and Russia. @frdfzi via @cyb3rops
• Threat actor targets security researchers with a fake RE tool — vx-underground flagged malware distributed as a reverse-engineering utility, aimed at compromising analysts who run it. A perennial reminder to detonate “tools of the trade” in isolation. @vxunderground
• Evooo1Bot — a new Linux botnet family that bolts the Mirai DDoS engine onto a more modular, capable framework and turns compromised routers into traffic-relay nodes. BleepingComputer
• ChainDrop worm hits the npm supply chain — a self-propagating worm in npm packages is reported to evade standard defenses, the latest in a steady run of registry-borne supply-chain threats. The Register
• Arrests over €30M bank fraud exploiting a service-provider flaw — Germany’s BKA charged three suspects in Europe and Brazil’s federal police arrested four more over a scheme that abused a vulnerability at a payment service provider to drain Commerzbank customers’ accounts, affecting German and Brazilian victims. The Record, BleepingComputer

Breaches

• RingCentral breach data hits Have I Been Pwned — roughly 1.6M unique email addresses, plus names, physical addresses, and phone numbers, were dumped following the July social-engineering intrusion tied to ShinyHunters (earlier coverage). The Register

AI & Model Security

• GLM-5.3 (ZAI ZCode) system prompts and tooling leaked — a dump of nearly 400k files reportedly includes the full system prompts, tools, and skills for the model, along with ZCode credentials — useful for anyone studying agentic tool wiring and the attack surface it exposes. @elder_plinius via @thegrugq
• Study pushes back on “autonomous AI research is imminent” claims — in work with Princeton and the UK AI Security Institute, agents built on Claude Opus 4.8 and GPT-5.6 Sol were given six days, $3,000 in credits, and GPUs to write research papers; NeurIPS authors rated the output “Reject.” Models handled the engineering but failed at research judgment and abandoning dead ends. The Decoder
• OpenAI’s “Computer History” logs clicks and keystrokes — the Mac feature records clicks, keystrokes, and app switches into a searchable timeline for ChatGPT and Codex, stored locally as unencrypted Markdown. A notable data-exposure surface on any endpoint where it’s enabled. The Decoder
• Qwen 3.8-27B released under Apache 2.0 — Alibaba’s Qwen team shipped a dense, natively multimodal 27B open-weight model with 262K context, runnable locally in ~17GB RAM via Unsloth GGUFs — attractive for local, air-gapped, and agentic security tooling. The Decoder