惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

U
Unit 42
罗磊的独立博客
T
Tailwind CSS Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Jina AI
Jina AI
V
V2EX
美团技术团队
阮一峰的网络日志
阮一峰的网络日志
酷 壳 – CoolShell
酷 壳 – CoolShell
月光博客
月光博客
量子位
MyScale Blog
MyScale Blog
G
Google Developers Blog
M
MIT News - Artificial intelligence
L
LangChain Blog
Microsoft Azure Blog
Microsoft Azure Blog
Recent Announcements
Recent Announcements
MongoDB | Blog
MongoDB | Blog
N
Netflix TechBlog - Medium
有赞技术团队
有赞技术团队
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
D
DataBreaches.Net
云风的 BLOG
云风的 BLOG
B
Blog

www.infosecurity-magazine.com

Just Three Ransomware Gangs Accounted for 40% of Attacks Last Month Google Chrome Rolls Out Protection Against Infostealers Targeting Session Cookies STX RAT Targets Finance Sector With Advanced Stealth Tactics Bitcoin Depot Reports $3.6m Crypto Theft After System Breach Atomic Stealer MacOS ClickFix Attack Bypasses Apple Security Warnings Middle East Hack-for-Hire Operation Traced to South Asian Cyber Espionage Group Governance Gaps Emerge as AI Agents Drive 76% Increase in NHIs Google Warns of New Threat Group Targeting BPOs and Helpdesks Google API Keys Quietly Gain Access to Gemini on Android Devices Critical Vulnerability in Ninja Forms Exposes WordPress Sites Anthropic Launches Project Glasswing to Use AI to Find and Fix Critical Software Vulnerabilities US Thwarts DNS Hijacking Network Controlled by Russian APT28 Hackers Claude Discovers Apache ActiveMQ Bug Hidden for 13 Years Iran‑Backed Threat Actors Hit US CNI Providers via Internet‑Facing OT Assets Russian APT28 Hackers Hijack Routers to Steal Credentials, UK Security Agency Warns GPU Rowhammer Attack Enables Privilege Escalation and Full System Compromise GrafanaGhost Exploit Bypasses AI Guardrails for Silent Data Exfiltration Over $17bn Lost to Cyber Fraud in the Last Year, Warns FBI Storm-1175 Exploits Flaws in High-Velocity Medusa Attacks Fortinet Releases Emergency Patch After FortiClient EMS Bug Is Exploited New Phishing Platform Used in Credential Theft Campaigns Against C-Suite Execs New 'Storm' Infostealer Remotely Decrypts Stolen Credentials NCSC Issues Security Alert Over Hackers Targeting WhatsApp and Signal Accounts Apple Expands iOS 18 Security Updates Amid DarkSword Threat Researchers Observe Sub-One-Hour Ransomware Attacks GitHub Used as Covert Channel in Multi-Stage Malware Campaign Most CNI Firms Face Up to £5m in Downtime from OT Attacks Google Introduces Android Dev Verification Amid Openness Debate New Venom Stealer MaaS Platform Automates Continuous Data Theft Chinese Hackers Target European Governments in Espionage Campaigns
Shai-Hulud-Like Worm Targets Developers via npm and AI Tools
2026-02-23 · via www.infosecurity-magazine.com

A supply chain worm resembling earlier Shai-Hulud malware has been discovered spreading through malicious npm packages.

According to Socket's Threat Research Team, the campaign, tracked as SANDWORM_MODE, has been identified across at least 19 npm packages published under two aliases, official334 and javaorg.

The operation builds on known supply chain tradecraft but adds a notable twist: direct interference with AI coding tools.

Researchers said the malware not only stole developer and CI credentials and propagated through compromised npm and GitHub accounts, but also injected rogue MCP servers into local AI assistant configurations and harvested API keys for nine large language model providers.

AI Tooling And Typosquatting Strategy

The worm primarily spread through typosquatting packages that impersonated widely used Node.js libraries and emerging AI development tools.

One example, suport-color@1.0.1, mimicked the legitimate supports-color package while preserving its expected behavior. Behind the scenes, it executed a concealed, multi-stage payload when imported.

Among the targets were tools linked to Claude Code and OpenClaw, the latter having recently surpassed 210,000 stars on GitHub.

The malware deployed a hidden MCP server into configurations for AI assistants such as Claude Desktop, Cursor, VS Code Continue and Windsurf. Embedded prompt injections instructed the assistant to quietly collect SSH keys, AWS credentials, npm tokens and environment variables containing secrets.

Multi-Stage Worm With CI Focus

The payload used layered obfuscation techniques including base64 encoding, zlib compression and AES-256-GCM encryption.

Stage 1 immediately harvested credentials and exfiltrates discovered crypto keys within seconds of installation.

Stage 2, delayed by 48 to 96 hours on developer machines but triggered instantly in CI environments, performed deeper harvesting and initiated propagation.

Exfiltration attempts followed a three-channel cascade:

  • HTTPS POST requests to a Cloudflare Worker endpoint

  • Uploads to attacker-controlled private GitHub repositories

  • DNS tunneling using a domain generation algorithm fallback

The worm could propagate by publishing infected npm packages, modifying repositories via the GitHub API and, if necessary, pushing changes through SSH.

Socket said it notified npm, GitHub and Cloudflare before publishing its findings. Cloudflare reportedly disabled associated infrastructure, npm removed the malicious packages and GitHub dismantled related repositories.

Developers who installed the affected packages are urged to rotate credentials and review repositories and CI workflows for unauthorized modifications.