Brief #122: CrowdStrike npm Attack, SonicWall Cloud Breach, ChatGPT Zero-Click

Nikoloz Kokhreidze

Nikoloz Kokhreidze

9 min read

Ransomware payments dropped to $115K median but hit 44% of all breaches. Cybersecurity pros can't find jobs despite 10+ years experience.

mandos brief cybersecurity newsletter by nikoloz kokhreidze

Happy Sunday!

In this week's brief:

  • CrowdStrike's npm packages got hit by the Shai-Hulud supply chain attack, affecting over 526 packages and showing how sophisticated these attacks have become
  • 71.7% of workplace AI tools are classified as high or critical risk, with most enterprise data flowing to risky platforms rather than secure alternatives
  • Experienced cybersecurity professionals are struggling in an unprecedented job market, with some reporting over 1,000 applications without success

A quick note before we dive in.

Industry News

Shai-Hulud Supply Chain Attack Targets CrowdStrike npm Packages

  • Multiple npm packages published under CrowdStrike's account were compromised in an ongoing supply chain attack called "Shai-Hulud" that previously affected tinycolor and 40+ other packages. The attack uses identical malware that scans systems for tokens/credentials and exfiltrates sensitive data.

  • The malware evolved through seven distinct versions, becoming increasingly stealthy and efficient. It self-propagates by stealing credentials to modify and republish packages, adding a malicious bundle.js file that executes automatically via postinstall scripts.

  • Over 526 packages were affected, with more than 100 packages compromised in a single burst targeting CrowdStrike. Organizations should rotate npm tokens, audit environments that installed affected packages, and monitor for unusual publish events.

ShadowLeak: Zero-Click Attack Exfiltrates Sensitive Data via ChatGPT's Deep Research Agent

  • Researchers discovered a zero-click vulnerability in ChatGPT's Deep Research agent that allows attackers to craft malicious emails that trick the agent into leaking sensitive inbox data with no user action required or visible indication.

  • Unlike previous attacks requiring client-side image rendering, this service-side exfiltration occurs entirely from OpenAI's cloud environment, making it invisible to traditional enterprise security controls and applicable across multiple data connectors beyond Gmail.

  • The attack used sophisticated social engineering techniques embedded in HTML emails to bypass the agent's safety restrictions, achieving 100% success rate by encoding stolen data in Base64 format before exfiltration.

SonicWall Urges Password Resets After Cloud Backup Breach Affecting Under 5% of Customers

  • SonicWall detected suspicious activity targeting their cloud backup service, with unknown threat actors accessing encrypted firewall configuration files for less than 5% of customers through a series of brute-force attacks.

  • While credentials within the compromised files were encrypted, the files contained information that could help attackers potentially exploit the related firewalls, prompting SonicWall to urge immediate password resets.

  • The company recommends affected customers take urgent containment actions including limiting WAN access, disabling management services and VPN access, resetting passwords, and importing fresh preferences files provided by SonicWall with randomized credentials.

Member-Only Content

Join Mandos to Continue Reading

Get instant access to this article and the Mandos Brief - your weekly 10-minute security leadership update.

Already a member? Sign in

Nikoloz Kokhreidze

Share With Your Network

Check out these related posts