Research Saturday

The wrong kind of access.

Dan Lisichkin⁠, AI Security Researcher from ⁠Pillar Security⁠, discussing their work on "A WIF Of Fresh Access, How a GitHub Issue on Gemini-CLI Led to GCP Project Compromise." Researchers at Pillar Security uncovered a vulnerability in Google’s Gemini CLI workflows that could allow an attacker to turn a malicious GitHub issue into remote code execution and gain elevated access to a Google Cloud Platform (GCP) project. The exploit chain combined a misconfigured tool allowlist that allowed unrestricted command execution with exposed Workload Identity Federation credentials and overly broad service account permissions, potentially enabling attackers to impersonate privileged accounts. Google promptly addressed the issues by tightening tool restrictions, blocking the agent’s access to credential files, and limiting service account impersonation permissions, highlighting the risks of connecting AI agents to untrusted input and sensitive infrastructure.

Contents may be malicious.

⁠Adam Meyers⁠, Head of Counter Adversary Operations at ⁠CrowdStrike⁠, is discussing their work on "PhantomRaven, An LLM-Generated Information Stealer Developed for Bug Bounty Hunting." PhantomRaven, a JavaScript-based information stealer distributed through malicious npm packages by a financially motivated threat actor posing as a bug bounty hunter. The malware targets system information and continuous integration and continuous deployment (CI/CD) environment variables, likely seeking credentials, with analysis suggesting its code was generated using a large language model. The report explores how AI-generated tools may lower the barrier to cybercrime and outlines steps organizations can take to mitigate risks, including restricting package installation scripts, using private npm registries, and monitoring dependencies.

Play to win, pay to lose.

Today we are joined by Jean-Pierre Mouton, Senior Threat Intelligence Consultant at GuidePoint Security, discussing their work on "How Play Achieves Encryption." Play ransomware, also known as PlayCrypt, continues to target organizations across multiple sectors using a consistent double-extortion playbook that combines data theft with widespread encryption. A recent investigation details how the group gained access through a SonicWall VPN, moved laterally using tools such as Mimikatz and PsExec, exfiltrated sensitive data, and used the victim’s own SentinelOne uninstallation utility to disable endpoint protection. The findings highlight several behavioral indicators defenders can monitor, including tool staging through SYSVOL and SystemBC for command and control, event log clearing, and suspicious WinSCP activity.

An apple a day, a phish away.

Today we are joined by ⁠Ensar Seker⁠, VP of Research and CISO at ⁠SOCRadar⁠, discussing their work on "Exposing AnonyMousKIT: AI-Powered PhaaS Supply Chain." An investigation into AnonyMousKIT reveals an AI-powered Phishing-as-a-Service platform designed to steal Apple credentials and disable Activation Lock on stolen devices. The platform uses email, SMS, WhatsApp, and AI-driven voice calls to impersonate Apple Support, with researchers uncovering a broader ecosystem spanning 506 domains, 168 storefront brands, and 30 backend installations. Despite its sophisticated social-engineering capabilities, basic coding flaws exposed extensive operational logs and revealed the shared infrastructure, developer activity, and reseller network behind the criminal operation.

All about that proxy.

Today we are joined by ⁠Dr. Renée Burton⁠, VP of Threat Intelligence at ⁠Infoblox⁠, discussing their work on Lurking Lizard, "Fake Installers, Fake Reviews, Fake Services – Real Proxies, Real." The research uncovers Lurking Lizard, a threat actor that has operated since at least 2022 by using fake software installers, VPNs, and lookalike domains to secretly turn victims’ devices into residential proxy nodes. Researchers identified more than 230 related domains and connected seemingly separate campaigns—including fake 7-Zip, downloader tools, and WireVPN—through shared infrastructure, tracking URLs, deployment patterns, and APIs. The investigation suggests the actor runs an end-to-end proxy operation, recruiting compromised devices and then monetizing their bandwidth through proxy services and fake review sites, with WireVPN appearing to be the latest evolution of the campaign.