PhantomRaven: LLM-Assisted Information Stealer Targets Bug Bounty Hunters

CrowdStrike's Counter Adversary Operations team has identified a novel JavaScript-based information stealer named PhantomRaven, allegedly created with the aid of large language models (LLMs). The malware is distributed by a threat actor who operates as a bug bounty hunter, leveraging the npm package registry to spread their malicious tool. This development highlights the increasing accessibility of sophisticated malware creation through AI assistance, potentially lowering the barrier to entry for malicious actors.

The threat actor behind PhantomRaven is a self-proclaimed bug bounty hunter who has been active since November 2022. Publicly available information suggests they have successfully collected bounties from numerous organizations across the technology, retail, and hospitality sectors through various bug bounty platforms. CrowdStrike's analysis indicates that PhantomRaven is likely used exclusively by its creator to discover vulnerabilities and secure bug bounty rewards, rather than being sold on underground markets.

CrowdStrike's managed detection and response (MDR) service encountered PhantomRaven within two npm packages: transform-jsbi-to-bigint and sort-imports-es6-autofix. Both packages were published under npm usernames (jpdhellonpm1 and jpd15) that appear to be controlled by the same individual, based on shared naming conventions and the inclusion of the threat actor's initials.

The infection vector relies on typosquatted npm packages that initially contain benign code. However, these packages are configured to fetch a remote dependency from attacker-controlled infrastructure via an HTTP URL. When a developer installs such a package, the npm client downloads and executes the PhantomRaven payload, often through a preinstall script.

This distribution method was partially mitigated by a June 2026 update to npm's package manager. Newer versions of npm now warn users about preinstall scripts in dependency packages and require explicit approval before execution, thus hindering the automatic deployment of malware like PhantomRaven. However, older versions or instances where users bypass these warnings could still be vulnerable.

Further investigation revealed the threat actor's attempts to publish malicious code on the Python Package Index (PyPI). While their efforts were flagged by PyPI administrators who recognized the intent to create an information stealer, associated Python files containing similar functionality to PhantomRaven were discovered. This suggests the actor is exploring multiple avenues for malware distribution across different ecosystems.

The PhantomRaven malware is designed to steal credentials and sensitive data from infected systems. Its use of LLMs for development is inferred from code characteristics such as verbose comments, placeholder code, and specific token-analysis patterns, which are common artifacts of AI-generated code. The malware also incorporates evasion techniques to avoid detection by security solutions.

The emergence of PhantomRaven underscores a growing trend where AI tools are being leveraged to develop and deploy malware. While the immediate threat appears contained to the bug bounty hunting community, the underlying technology and techniques could be adopted by more malicious actors for broader cybercrime operations.