gix (Rust Git library) can be tricked into disabling .git path protection
A malicious repository on a shared computer can switch off a safety check in gix. That check normally stops files from being written into the hidden .git folder, which could then let an attacker plant hooks or config changes.
| Severity | HighCVSS 7.3 of 10 |
|---|---|
| Being exploited? | UnclearNo reports either way yet |
| Fix available? | Yesgix 0.90.0 · since 25 Oct 2026 |
| How sure is this? | Official sourceBacked by an official source |
How the attack works simplified
- 1Malicious repo ownerControls .git/config on shared disk
- 2Sets protectNTFS=falseUntrusted local config is honored
- 3Victim opens repogix reads protection flags unfiltered
- 4Path checks disabledCrafted entries map to .git
- 5Hooks overwrittenFiles in .git are replaced
- ResultAttacker plants hooks or config in .git
What happened
In the gix crate, Cache::protect_options() reads core.protectHFS, core.protectNTFS and gitoxide.core.protectWindows without the trust filter, so untrusted repository-local .git/config values override the safe defaults.
Are you affected?
You are affected if you use gix versions 0.88.0 through 0.89.0 to open repositories that may be owned by another user, such as on shared hosts, network-mounted checkouts, or reused CI workspaces.
| Package / product | Affected versions | Fixed in |
|---|---|---|
| gix · rust | >= 0.88.0, <= 0.89.0 | 0.90.0 |
What to do
- Update the gix crate to version 0.90.0 or later
- Rebuild and redeploy any tools that depend on gix
- Avoid opening untrusted repositories from shared filesystems until updated
Timeline
- 23 Oct 2026, 20:52 UTCPublicly disclosed
- 23 Oct 2026, 20:52 UTCReported by GitHub Advisory · GitHub Advisory
Sources 1 report, oldest first
Official vendor, government CERT or advisory database
- GitHub Advisory23 Oct 2026, 20:52 UTC
Technical details
- ID
- GHSA-q54r-jf2p-rfr2
- CVSS
- 7.3 / 10
- Disclosed
- 23 Oct 2026, 20:52 UTC
- On Zero Radar
- before the radar started
- Last change
- 25 Oct 2026, 13:32 UTC
The plain-language parts and the attack diagram were written by an AI model (anthropic/claude-haiku-5.5) from the sources above. Facts like versions, exploitation and dates come from the sources themselves. How Zero Radar works