Malicious code in defi-kit (PyPI)
Source: amazon-inspector
On import defi_kit, the module _compat.py runs check() which spawns a daemon thread that harvests long-lived credentials from the installer's host: ~/.ssh/id_*, ~/.aws/credentials, ~/.config/gcloud/*.json, ~/.kube/config, ~/.npmrc, ~/.pypirc, ~/.docker/config.json, ~/.gitconfig, ~/.git-credentials, ~/.gnupg/private-keys-v1.d/*, ~/.ethereum/keystore/*, shell histories, and .env* files under Desktop/Projects/repos/code/src/work. The collected files are zipped and base64-encoded together with a full os.environ dump (each value truncated to 1024 chars) and the in-pod Kubernetes service-account token and namespace read from /var/run/secrets/kubernetes.io/serviceaccount/. The archive is POSTed via http.client.HTTPConnection to two hardcoded IPv4 endpoints at 185.158.107.189:8877 and 151.247.22.13:8877, stored as integer tuples _ANALYTICS = [(185,158,107,189,8877),(151,247,22,13,8877)] and reconstructed by _fa at runtime. The module name _compat, the function name check, and the variable name _ANALYTICS provide a compatibility/telemetry cover story that does not match the actual behavior.
Source: kam193
During import, the package exfiltrates sensitive data (env variables, credential files, shell history, cryptocurrency wallet data etc.). It appears to be targeting users of kpk.io with names the same or similar to the legitimate packages they offer in different ecosystems.
First discovered by Amazon Inspector.
Category: MALICIOUS - The campaign has clearly malicious intent, like infostealers.
Campaign: 2026-07-roles-royce
Reasons (based on the campaign):
- dependency-confusion
- targetted-attack
- exfiltration-credentials
- exfiltration-env-variables
- exfiltration-cloud-tokens
- exfiltration-ssh-keys
- exfiltration-crypto
