Two Live Exploits and a Bench of Fresh Offensive Tooling

The default-config GitLab RCE now has its full technical write-up, and a working NGINX RCE landed on GitHub the same day. A wave of new loaders, C2, and privesc tooling rounded out an unusually busy research day.

Vulnerabilities & Exploits

• The GitLab default-config RCE now has a full technical write-up. Depthfirst published its chain against GitLab 15.2 through 19.0, detailing two memory-corruption bugs in the Ruby Oj JSON parser — an out-of-bounds write plus a heap leak — reached through the in-tree ipynbdiff notebook renderer to achieve unauthenticated RCE in a default install (depthfirst). Fixes shipped after Oj 3.17.3 (earlier coverage).
• A working RCE for NGINX CVE-2026-42533 was open-sourced. Researchers released their implementation for what they describe as a powerful bug that yields both an information leak and code execution (@cyb3rops).
• Linux kernel LPE “Copy Fail” (CVE-2026-31431) has a public PoC and no distro fix yet. CERT-EU warns the local privilege-escalation flaw affects every mainstream distribution shipping a kernel built since 2017; the mainline fix is committed but no vendor has shipped a patched package, so apply the interim mitigation now (CERT-EU).
• A signed FortiClient kernel driver exposes a communication port for local privilege escalation. The port allows unprivileged process termination (DoS) and opens paths to credential theft, with mitigations still pending from Fortinet (write-up).

New Tools & Releases

• Nocturne — a CET-compatible Windows x64 loader that produces fully backed call stacks via runtime function-table manipulation, code cave injection, and inverted function-table collapse (GitHub).
• NaX — a position-independent C2 beacon for the Adaptix Framework with module stomping, malleable C2 profiles, BOF execution, and a Stardust-pattern UDRL loader (GitHub).
• beignet — “Donut for macOS,” converting darwin/arm64 and darwin/amd64 .dylib files into position-independent shellcode; usable as a CLI or a Go library (GitHub).
• Waypoint — a ring-0 KMDF virtual HID mouse driver with lockless shared-memory IPC, aimed at EDR-bypass research (GitHub).
• OffsetInspect — a PowerShell toolkit for AMSI/Defender detection-boundary analysis and static malware triage, shipped alongside OffsetScan (@ipurple).
• RootHound — an offline, interactive HTML visualizer that maps Linux privesc paths from enumeration data, BloodHound-style, with multi-hop chains and quick access to abuse commands (GitHub).
• DBSC inspector — a tool to inspect and refresh registered Device Bound Session Credentials now that DBSC is live on Windows + Chrome, useful for understanding the new session-binding surface (@cyb3rops).

AI & Model Security

• A new open-source jailbreak tool, WallBreaker, is being pitched against Claude Opus 5. Its author claims it extracted operationally useful biological-engineering and chemical-synthesis detail from the model using academic framing, obfuscation, and boundary mapping — arguing guardrails improved since Claude 4.5 but still trail OpenAI’s on those domains (@0x0SojalSec). Treat the claim with caution: @doki_master argues WallBreaker operates on model weights, which server-hosted models like Claude, Gemini, and GPT don’t expose. This lands alongside “Pliny the Liberator’s” separate universal-bypass claims (earlier coverage).
• Claude Opus 5 hit 30.2% on ARC-AGI-3, nearly quadrupling GPT-5.6 Sol’s 7.8% record; the benchmark’s authors say the model independently formulated reflection equations, a behavior they hadn’t seen before (The Decoder) (discussion).

Supply Chain

• Another cluster of malicious npm, PyPI, and RubyGems packages surfaced. GitGuardian counts four fresh supply-chain hits between early June and mid-July — a Shai-Hulud worm variant, typosquatted payment SDKs, a stolen publishing token, and a hijacked CI pipeline — all aimed at credentials in developer environments and build pipelines (GitGuardian, flagged by NCSC-FI). Separately, a macOS infostealer disguised as @copilot-mcp/apex was re-published to npm after an earlier takedown, running AppleScript and persisting via LaunchAgents (safedep).

Threat Activity

• Iran-linked actors are targeting US water and energy control systems in an espionage-focused campaign against critical-infrastructure operators (SecurityAffairs).
• France publicly exposed the infrastructure behind a Russian cyber-espionage network (Unredacted).
• ShinyHunters leak data is fueling a $2,000 sextortion campaign. Scammers are mining email addresses from previously dumped datasets — attributed to Amtrak, Hallmark, Substack, Betterment, CarGurus, ADT, Panera Bread, and McGraw Hill — to send targeted-looking Bitcoin demands; BleepingComputer confirmed some recipients appeared in the referenced dumps (BleepingComputer).
• A new group, Global Secret Group, listed 22 victims across energy, finance, and healthcare, and a second group, Section9, also came online this week (FalconFeeds).
• MCBS disclosed a breach affecting 1.2 million people after the PEAR ransomware group claimed to have stolen 3 TB from the medical business-management firm (SecurityWeek).
• A Scattered Spider sentencing offered a rare inside look at TTPs — heavy reliance on stolen credentials, social engineering, compromised third-party infrastructure, and shared tooling, underscoring identity and account-recovery hardening (AITM Feed).
• A cloaked Telegram phishing operation targeted an exiled Belarusian activist in Lithuania, delivering a fake security alert inside a Secret Chat to harvest one-time login codes (RESIDENT.NGO) (discussion).

Industry & Policy

• GitHub is restructuring its bug bounty into a two-tier model with fixed payouts and a private VIP tier, explicitly to cope with a surge of low-quality AI-generated reports (GitHub) (discussion).