PhantomRaven , a dangerous information stealer written in JavaScript , is the focus of new research from CrowdStrike , which uncovers an unusual threat actor: a self-proclaimed bug bounty hunter who likely used LLM (Large Language Model) to develop his malware. The case highlights how artificial intelligence is transforming the cyberthreat landscape, dramatically reducing the time and expertise required to create sophisticated malware .
According to CrowdStrike Counter Adversary Operations analysis , the attacker is distributing PhantomRaven via the npm package registry , utilizing slopsquatting and typosquatting techniques . The campaign was first detected by Koi Security and DCODX in late October 2025 , when over 100 malicious packages were discovered in the npm registry . Their goal was to steal authentication tokens , CI/CD secrets , and GitHub credentials from developers’ machines.
What makes the case particularly interesting is the perpetrator's motive: he doesn't appear to be selling the stolen data on underground marketplaces, but rather using it as "evidence" to claim rewards from official bug bounty. It's a strategy that blurs the lines between legitimate security research and cybercrime.
See also: GitHub reduces Public Bug Bounty rewards and moves the biggest rewards to a private program
How PhantomRaven works in the npm ecosystem
The software supplychain attackimplemented by PhantomRaven is technically sophisticated. The malicious packages that the attacker uploads to npm do not contain the malware; instead, they retrieve remote dynamic dependencies (RDDs) from an external server during installation. This technique allows the packages themselves to evade detection by security tools that scan the code statically.
Once the package is installed, PhantomRaven is activated and starts scanning the development environment. It collects email addresses , information about the CI/CD environment , system fingerprint (including public IP address ), and sends all data to a server under the attacker's control. In addition, the stealer collects runtime details , date and time, usernames, and email addresses from Git and npm configurations .
Particularly concerning is PhantomRaven 's ability to extract CI/CD environment variables from platforms like GitHub Actions , GitLab CI , Jenkins , and CircleCI . These variables often contain sensitive API keys , deployment credentials , and secrets that can give access to entire production infrastructures.
PhantomRaven and the use of AI: A new era for malware
CrowdStrike concluded that the perpetrator used LLM to write PhantomRaven based on specific technical clues: verbose comments , placeholder code that has not been stripped, and statistical token parsing patterns that are characteristic of AI-generated text. These clues suggest that the perpetrator is not an experienced programmer, but leveraged AI technology to compensate for his lack of technical expertise.
See also: Browser Company: Bug Bounty Program for Arc Browser
The threat actor has been active since November 2022 and claims to have collected bug bounty from at least 9 entities in the technology, retail, and hospitality sectors. In August 2025, the perpetrator claimed to have discovered a Remote Code Execution (RCE) via a malicious npm package he published, explaining that he managed to execute preinstall script on a target machine.
Online identities associated with the same operation include jpd12 , jpd13 , npmhell , npmpackagejpd , npmtestdharsh , jpdhackerone11 , and packagedharsh . At least two different npm accounts held by the perpetrator have now been deactivated. There is also evidence that the perpetrator attempted to upload packages to the Python Package Index (PyPI) with information stealer code that bears similarities to PhantomRaven .
CrowdStrike notes that it has not seen the stolen data appear on stealer log shops or underground forums, which reinforces the assumption that the perpetrator is using PhantomRaven solely to identify bug bounty. This strategy — using malware to prove the existence of a vulnerability and then claiming a reward — is a serious violation of the terms of legitimate bug bounty and may constitute a criminal offense.
Protection from PhantomRaven and similar threats npm
The threat posed by PhantomRaven highlights the need for stricter security practices in software development. Organizations and developers should implement Software Composition Analysis (SCA) tools that check dependencies before installation, use lockfiles to prevent automatic rollback of unexpected versions, and actively monitor network connections made during package installation.
See also: Apple Bug Bounty: Maximum reward $2 million, $35 million to date
🔒 Protect your privacy with Proton VPN
Swiss VPN from the creators of Proton Mail — strict no-logs policy, strong encryption, and built-in NetShield that blocks ads, trackers, & malware.
- ✔ No-logs, based in Switzerland (except 14-Eyes)
- ✔ NetShield: blocks ads, trackers & malicious domains
- ✔ Covers all devices — free version available
The link is an affiliate link — SecNews may receive a commission at no additional cost to you. It does not affect the independence of our article writing.
Additionally, it is critical to apply the principle of least privilege to CI/CD pipelinesso that even if a malicious package is executed, it does not have access to sensitive secrets and credentials. Using secret scanning tools and regularly rotating API keys can also limit the damage in the event of a breach. GitHub Actions and other CI/CD offer built-in features for protecting secrets that should be fully utilized.
The phenomenon of using AI to create malware is not new, but the PhantomRaven case is one of the most well-documented examples to date. As LLMs become increasingly accessible and powerful, we can expect to see more threat actors without deep technical training develop sophisticated attack tools. This poses new challenges for the cybersecurity community, which must adapt its detection and response methods. According to The Hacker News, CrowdStrike continues to monitor the perpetrator and work with relevant authorities.
