Security context
What an agent needs to avoid regressing past fixes and find the next vuln in this repo.
pyg-team/pytorch_geometric
master @ 1f0661c
3
Fixes
0
CVEs
HIGH
Peak severity
Highlights
Denial of Service: 2 prior fixes. Scrutinize any change in this area.
torch_geometric/data/extract.py: most-fixed (1 issue). Treat as high-risk during review.
1 high-severity fix in this history; regressions here are high-impact.
Recurring patterns
The bug types that recur here, drawn from past fixes, not open vulnerabilities.
Path Traversal: Unsafe archive extraction (Tar Slip) can allow attackers to write arbitrary files to the filesystem, potentially achieving remote code execution if key files are overwritten. This was addressed by introducing extraction filters.
Denial of Service: An infinite loop during negative triangle sampling allows malicious or malformed graphs to hang the execution thread indefinitely due to an improperly updated residual array.
Denial of Service: Integer overflows during edge index sorting and coalescence operations can lead to crashes or undefined tensor operations under extremely large graph structures.