GitHub hosts malware too

A search for a well-known app that lands on github.com feels like arriving at the source. It is not. GitHub is a hosting provider, not a curator: anyone can open an account, publish a repository, and serve files from one of the most trusted domains on the web. Attackers know this, and they build on it. Repositories, organization profiles, and free *.github.io websites are all in play.

Brand impersonation is routine

The fakes do not just copy the marketing. They borrow the brand, and both closed-source and open-source projects get copied:

Read these captures, but do not click inside them. The Wayback Machine preserved the pages; the download buttons still resolve to live attacker infrastructure. Treat every link inside them as hostile.

These two are not outliers. In July 2026, Arctic Wolf Labs documented one actor behind at least 292 repositories impersonating software vendors, security companies, and developer tools, Arctic Wolf included. Every repo concealed a download link that hopped through a GitHub Pages address to an outside domain serving the BoryptGrab infostealer, the same shape as the CCleaner fake above. Arctic Wolf’s write-up.

Impersonation is not the only shape. SwiftClean borrowed no brand at all: it invented one, announced itself on Reddit as a “free and open-source MIT” cleaner for Windows and macOS, and shipped a password-protected zip while its repository held no source code. Manual analysis confirmed the payload malicious, the only confirmed malware in this catalog so far. Same host, same tricks, no impersonation required.

Why attackers pick GitHub

The same properties that make GitHub the default home of real software make it a comfortable home for the fake kind:

Even the history is manufactured

The freshness signal can be counterfeited too. Another repository dressed as Notepad++, abovedaemons-59/Notepad, showed dozens of commits, all authored by github-actions[bot], each with a plausible generated message and an issue number like “#4268” implying thousands of issues that do not exist. Open any of them and the diff is one line: a “Last updated” timestamp in the README, bumped by a few hours. The repository reports a fresh last-commit time and a healthy commit count without ever changing anything. Archived commit history.

They are not hiding

None of this took any finding. GitHub’s own search box returns the same shape: a query for free download full version crack lists repositories named exactly that, like bludmoon43/aomei-partition, titled “AOMEI Partition Assistant Crack | Free Download Full Version”, a README of keywords and one “Click Here” link leaving GitHub, still up twenty months after it was created. Google finds them just as fast: site:github.com free cracked full download official returns pages of the same.

Which means detection is not the bottleneck. GitHub indexes all of this already; a five-second keyword search available to anyone is available to it. When fakes this visible stay up for months at a time, the gap is priorities, not capability.

The tells

No single sign is proof, and honest projects will trip one or two. Together they draw the shape:

What to do instead

The takeaway

GitHub the company removes reported abuse, and pre-screening a firehose of new accounts and repositories is genuinely hard; it never claimed to. Still, the fakes above are not the hard part of the firehose: they are the part a five-second keyword search finds. A github.com URL tells you where a file is hosted, not who vouches for it. The rest of the checklist lives at is this a good cleaner?.

Last reviewed