PCPJack: Credential-Stealing Worm Exploits 5 CVEs to Spread Across Cloud Infrastructure
Security researchers at SentinelOne have detailed a new credential theft campaign they're calling PCPJack. It's a modular worm that goes after exposed cloud services — Docker, Kubernetes, Redis, MongoDB, RayML — and spreads by exploiting known vulnerabilities.
The attack chain starts with a bootstrap shell script. That script preps the environment, downloads next-stage Python tooling, terminates any TeamPCP processes already running on the box, and then settles in for the long haul. It even installs Python if it's not there already.
Five CVEs fuel the spread: CVE-2025-55182, CVE-2025-29927, CVE-2026-1357, CVE-2025-9501, and CVE-2025-48703. All are known flaws in the target platforms. If you're patched, you won't get owned this way. That's the tl;dr.
What's interesting is the relationship to TeamPCP, a threat actor that made noise late last year using similar TTPs — exploiting React2Shell and misconfigs in cloud services. PCPJack actively removes TeamPCP artifacts from compromised hosts. When it reports home, it even includes a "PCP replaced" field in its C2 traffic, essentially saying "yep, we handled the squatters." SentinelOne's Alex Delamotte noted this implies the actor was specifically focused on clearing out competitors rather than just opportunistic cloud exploitation.
The credential haul is broad: cloud services, container environments, developer tools, productivity apps, financial platforms. The C2 channel is Telegram — simple, disposable, and unlikely to get flagged by your average perimeter security stack.
One thing that stands out: PCPJack doesn't deploy cryptocurrency miners. TeamPCP did. Either the operator has a different monetization plan, or they're planning to sell the stolen credentials instead of turning cycles into cash. That part isn't clear yet.
The propagation logic pulls target lists from Common Crawl's parquet archives — so it's automating reconnaissance on a massive dataset to find exposed services. The check.sh script handles OS detection and picks the right Sliver binary, then queries IMDS endpoints, Kubernetes service accounts, and Docker instances for credentials tied to Anthropic, Digital Ocean, Discord, Google API, Grafana Cloud, HashiCorp Vault, and others.
Bottom line: if your cloud services are internet-facing and unpatched, you're in someone's crosshairs. The fact that one actor is actively kicking another off compromised hosts tells you there's real money in this. Not script kiddie stuff — organized, deliberate credential harvesting at scale.
Patch the five CVEs. Lock down IMDS access. Monitor for unexpected Python spawning and outbound Telegram traffic. That's the stack.
---
*Source: [The Hacker News — PCPJack Credential Stealer Exploits 5 CVEs to Spread Worm-Like Across Cloud Systems](https://thehackernews.com/2026/05/pcpjack-credential-stealer-exploits-5.html)*