CVE-2026-111865: THU-MIG yoloe AutoBackend Model Loading autobackend.py eval eval injection
A security flaw has been discovered in THU-MIG yoloe up to 40cd606cabdbe2b566d6f14a6b162c89206e9a1b. This affects the function eval of the file ultralytics/nn/autobackend.py of the component AutoBackend Model Loading. The manipulation of the argument imgsz/names/kptshape results in improper neutralization of directives in dynamically evaluated code. The attack requires a local approach. The exploit has been released to the public and may be used for attacks. This product does not use versioning. This is why information about affected and unaffected releases are unavailable. The project was informed of the problem early through an issue report but has not responded yet.
Affected Software
Event History
Frequently Asked Questions
What level of access does an attacker need?
Exploitation requires local access and low privileges. No user interaction is required.
Which inputs are involved in the injection path?
The issue is triggered by manipulating the imgsz, names, or kpt_shape arguments used by eval in ultralytics/nn/autobackend.py during AutoBackend model loading.
How can I determine whether my deployment is affected?
Check whether your code includes the affected eval usage in ultralytics/nn/autobackend.py and whether it is based on or includes commits up to 40cd606cabdbe2b566d6f14a6b162c89206e9a1b. The project does not use versioning, so affected and unaffected release information is unavailable.
Is exploit code available?
Yes. A public exploit has been released, increasing the likelihood that the flaw may be used in attacks.
What should be done if no vendor response or patch is available?
No patch or vendor response is reported. Limit local access and low-privilege access to systems that load models, and avoid processing untrusted values for imgsz, names, or kpt_shape through the affected model-loading path.