AIKIDO-2026-220941
torch is vulnerable to Out-of-bounds Write
55
Medium Risk
This Affects:
torchTL;DR
The mobile FlatBuffer loader reached through torch::load parses serialized module data without validating the offsets embedded in the buffer. Because untrusted offsets are dereferenced and combined with buffer pointers, a crafted FlatBuffer-format model can make the loader compute and access arbitrary memory addresses. Loading such a file can corrupt pointers and cause out-of-bounds reads and writes, at minimum crashing the process. The fix adds buffer verification and bounds checks so malformed modules are rejected before parsing.
Who does this affect?
You are affected if you are using a version that falls within the vulnerable range and your application loads FlatBuffer-format model data from untrusted sources via torch::load.
Background info
torch is vulnerable to Out-of-bounds Write in versions 0.0.1 - 2.0.1.
How to fix this
Upgrade the pytorch library to the patch version.
Are You Affected?
Connect your repositories to instantly see whether vulnerable or malicious packages exist in your codebase.
Free. No credit card required.
I consent to receiving marketing communications based on Aikido’s Privacy Policy.
SOC 2Compliant
ISO 27001Compliant