Security Check-in Quick Hits: North Korea’s $280M Social Engineering Masterclass, Claude Code Bypass Flaw, npm Supply Chain Malware, and Ransomware BYOVD Escalation
North Korea’s Most Terrifying Crypto Hack Yet – 6 Months of Patience Nets $280 Million from Drift Protocol
In a jaw-dropping display of advanced persistent threat tradecraft, North Korean actors (DPRK) executed what experts are calling one of the most sophisticated social-engineering campaigns in crypto history. Rather than relying on phishing or smart-contract exploits, the attackers spent six months building genuine trust with Drift Protocol contributors.
The operation began in Fall 2025 at major crypto conferences. Impersonators posing as a legitimate “quant trading firm” engaged in face-to-face meetings across multiple countries. They followed up with verifiable backgrounds, professional networks, and even deposited over $1 million of their own capital into an Ecosystem Vault. By early 2026, they had established months of normal business conversations via Telegram, attended working sessions, and shared routine repos – exactly the kind of collaboration that happens daily in the industry.
The kill shot? A known VSCode and Cursor vulnerability (flagged by the security community since late 2025) that enabled silent code execution simply by opening a malicious file. No prompts. No warnings. $280 million vanished instantly. Every trace – Telegram messages, malware artifacts – was wiped clean. The real victims never shook hands with actual North Koreans; the actors used cutouts and intermediaries for the in-person element.
Why this matters: This wasn’t a code exploit – it was a handshake exploit. Every founder, developer, and ecosystem partner now has to treat long-term relationship-building as a potential attack vector. The bug is patched, but the human vector remains wide open.
Critical Claude Code Vulnerability Lets Attackers Silently Bypass All Developer Security Rules
Anthropic’s Claude Code AI coding agent – used by hundreds of thousands of developers – contains a high-severity security bypass flaw that completely disables user-configured deny rules under specific conditions.
Security researchers traced the issue to bashPermissions.ts (lines 2162–2178). A performance optimization caps security analysis at 50 subcommands per shell command. When an attacker pads a command with more than 50 subcommands (joined by &&, ||, or ;) , Claude Code skips all deny-rule enforcement and falls back to a generic permission prompt. The result: malicious actors can silently steal credentials and compromise supply chains without triggering any configured safeguards.
The vulnerability was publicly detailed today by both Cybersecurity News and Adversa.ai. It highlights a growing risk in AI-powered developer tools: when security checks are treated as performance bottlenecks, attackers will weaponize that trade-off.
Immediate action for developers: Review all Claude Code workflows, limit command complexity where possible, and monitor for anomalous permission prompts. Anthropic has been notified; patches are expected soon. Until then, treat AI coding agents as high-privilege environments requiring extra scrutiny.
36 Malicious npm Packages Posing as Strapi Plugins Deliver Full Supply-Chain Compromise
The npm ecosystem took another major hit: 36 packages masquerading as legitimate Strapi plugins were discovered delivering malware through postinstall scripts. Attackers exploited Redis and PostgreSQL environments, stole credentials, and deployed backdoors that grant full user or CI/CD pipeline access.
Once installed, the malicious packages executed credential-harvesting routines and established persistent access. Forensic artifacts were cleaned up post-execution, making detection harder. The campaign targeted developers integrating Strapi (a popular headless CMS), turning routine dependency updates into silent compromises.
This incident joins a growing wave of supply-chain attacks aimed squarely at developer tooling and CI/CD pipelines. The speed – two backdoored versions of a related library (Axios) published within 39 minutes in a separate but related campaign – shows how quickly attackers can pivot once maintainer credentials are compromised.
Key takeaway for teams: Mandate signed dependencies, SBOM validation, and strict vetting of any npm package (especially those claiming to be “plugins” for popular frameworks). Postinstall scripts should be treated as automatic red flags unless explicitly reviewed.
Qilin and Warlock Ransomware Weaponize BYOVD to Slaughter EDR Before Attacks Even Begin
Two active ransomware families – Qilin and Warlock – are now routinely disabling endpoint detection and response (EDR) tools using Bring Your Own Vulnerable Driver (BYOVD) techniques before deploying their payloads.
Qilin side-loads a malicious DLL that leverages vulnerable kernel drivers to kill over 300 EDR-related drivers. Warlock takes a slightly different route, exploiting SharePoint environments while still relying on the same vulnerable drivers to bypass kernel-level protections. Both groups often delay ransomware execution to ensure defenses are fully neutralized.
The technique is devastatingly effective because it operates at the kernel level before most behavioral detection can react. Security teams are seeing longer dwell times and higher success rates for these campaigns.
Defender checklist: Audit kernel drivers for known vulnerable signatures, enable strict driver signing policies, and monitor for DLL side-loading attempts. Traditional EDR alone is no longer sufficient against these mature ransomware operations.
German BKA Pins REvil Ransomware Leader Behind 130 Attacks and €35M+ in Damage
Germany’s Federal Criminal Police Office (BKA) has publicly identified Daniil Shchukin (“UNKN”) as a key leader of the notorious REvil ransomware group. He is linked to at least 130 attacks in Germany alone, causing over €35.4 million in damages with €1.9 million paid in ransoms.
The identification marks a significant law-enforcement win against one of the most prolific ransomware operations of the past few years. REvil’s tactics – double extortion, data leaks, and high-profile targets – set the standard that many current groups still follow.
While arrests and identifications don’t always dismantle entire operations, they raise the personal risk for operators and provide intelligence for ongoing disruptions.
Bottom line for organizations: Treat ransomware as a persistent nation-state-level threat. Assume your data is already being shopped on leak sites and invest in offline backups, immutable storage, and rapid incident response capabilities.
Stay vigilant – the threat landscape moves fast, and today’s chatter becomes tomorrow’s headlines. Check back for the next Security Check-in Quick Hits.