Supply-chain attacks on open-source packages aren’t hypothetical anymore. Every few weeks there’s another story about a malicious package sneaking onto PyPI or npm — sometimes hiding behind a name that looks just close enough to something popular, sometimes just brand new and hoping nobody looks too closely before enough people install it.
I got curious how hard it would be to keep an eye on that myself, so I put together a small automated pipeline that watches both registries for newly published or updated packages around the clock. Most of what shows up isn’t worth a second look, so before anything gets a deeper scan, it has to clear a quick filter: is this package unusually popular, does its name look like it’s impersonating something well-known, or is it brand new with basically no track record? Those are the shapes that most real incidents actually take.
Anything that clears that bar gets run through an open-source malware scanner (GuardDog, built by Datadog) that looks for the kind of behavior that shows up in genuinely malicious packages — things like destructive file operations, code that spawns processes it has no business spawning, or code that goes out of its way to obscure what it’s actually doing.
A single scanner can trip over its own blind spots, though — the same behavioral shapes it’s watching for can show up by coincidence in a completely legitimate package that just happens to bundle a minified copy of some unrelated library. So before anything actually gets published here, a second, independently-built scanner (Socket) has to agree. Socket looks at a different mix of signals than GuardDog does — code inspection, package and maintainer metadata, and a maintained, human-reviewed list of confirmed malware — and only packages both tools flag make it onto this page. It’s been an interesting way to watch two differently-built tools agree, and just as often not.
This is a hobby project, not a security product — it won’t catch everything, and it’s one signal among many. But it’s been a fun way to get a little visibility into a corner of the open source ecosystem that attackers have been targeting heavily the past year.
Loading findings…