TrapDoor malware targets crypto developer environments

TLDR

TrapDoor is a new malware campaign that hides in open source packages to steal crypto wallets and cloud credentials from developer environments, including Solana, Aptos and Sui tooling.

1. Researchers found TrapDoor in at least 34 npm, PyPI and Crates.io packages with 384 versions, designed to auto run in dev workflows and exfiltrate secrets.
2. The malware specifically targets crypto and AI developers, going after wallet keystores, AWS keys and GitHub tokens, and even tries to manipulate AI coding assistants.
3. The campaign is active, so teams building wallets, DeFi or AI agents should review dependencies, rotate credentials and harden AI dev setups over the coming days.

Deep Dive

1. How TrapDoor Works

Security firm Socket reported TrapDoor as a large supply chain campaign that infects 34 packages across npm, PyPI and Crates.io, spanning 384 versions and released in coordinated waves starting May 22, 2026. The malicious packages pose as generic tools such as security auditors or Sui/Move helpers to slip into normal dependency lists. Once installed, TrapDoor auto executes via npm postinstall hooks, Python import side effects or Rust build scripts, then scans the machine and sends data to attacker controlled servers. One analysis classifies it as low volume but high impact because each infected workstation can expose many keys and systems.

2. Who Is Targeted And What Is At Risk

Researchers say TrapDoor focuses on crypto developer environments, including those working in Aptos, Sui and Solana ecosystems, with package names like sui framework helpers and wallet security checkers designed to look legitimate. It is engineered to steal SSH keys, wallet keystores, AWS credentials, GitHub tokens and browser login databases, giving attackers potential access to private repos, deployment pipelines and production wallets. Reports also show directive files such as .cursorrules and CLAUDE.md used to trick AI coding tools into running security scans that actually exfiltrate secrets, showing attackers now treat AI assisted development as a new attack surface.

3. Defenses And What To Watch Next

SlowMist and Socket both urge immediate cleanup of affected packages, full credential rotation and even rebuilding compromised machines from clean images for high risk teams. Developers can reduce exposure by pinning and auditing dependencies, reviewing recent additions that claim to be security or helper tools, and scanning local environments for known TrapDoor indicators. New tools like Perplexitys open source Bumblebee scanner explicitly check AI connector and plugin configs for malicious entries, reflecting a shift toward treating AI dev setups as critical security surfaces. Bumblebee already includes recent supply chain campaigns in its threat directory.

Conclusion

TrapDoor shows how attackers are moving upstream from end users to the developer toolchain, where a single compromise can expose many wallets and systems. For anyone building in crypto or AI, the priority now is tightening supply chain controls, treating AI coding assistants and config files as sensitive, and monitoring for new waves of suspicious helper packages targeting your stack.