Security researchers at CrowdStrike identified a threat actor known as PhantomRaven who posed as a bug bounty hunter. The attacker distributed malware created with the help of LLMs, masking it as legitimate libraries within npm repositories. Once installed, these packages executed a stealer to collect system data and gain unauthorized access to corporate resources.
The hacker used these compromised systems to identify vulnerabilities and submit reports to bug bounty programs. PhantomRaven has been active since 2022, claiming rewards from at least nine companies across the retail, hospitality, and technology sectors. The attacker also collaborated with platforms such as Bugcrowd, Intigriti, YesWeHack, HackenProof, and HackerOne.
CrowdStrike analysts noted that the stealer code contained verbose comments and statistical patterns typical of LLM-generated content. While the attacker attempted to expand operations to the PyPI repository, the request was rejected. Evidence suggests the stolen data was used solely for bug bounty submissions rather than being sold on dark web marketplaces.
The attack vector has been mitigated by updates to the npm package manager, which now blocks installation scripts in dependencies by default. CrowdStrike confirmed that they have remediated several incidents involving this actor. The case highlights the growing integration of AI tools into cyberattacks, ranging from reconnaissance to malicious code generation.
