Medical Sam Adapter
Monthly
Code injection in ImprintLab Medical-SAM-Adapter 0.1.0-alpha and 0.1.0 allows a local, low-privileged attacker to execute arbitrary code in the context of the user who loads a crafted model file. The flaw resides in the AutoBackend model-loading path (models/MobileSAMv2/ultralytics/nn/autobackend.py), where the imgsz, names, or kpt_shape metadata fields are passed into a dynamically evaluated expression without neutralization (CWE-95). Publicly available exploit code exists, and the project has not responded to the issue report; exploitation requires local access and victim interaction to load the malicious model, limiting impact to Low confidentiality, integrity, and availability.
Code injection in ImprintLab Medical-SAM-Adapter 0.1.0-alpha and 0.1.0 allows a local, low-privileged attacker to execute arbitrary code in the context of the user who loads a crafted model file. The flaw resides in the AutoBackend model-loading path (models/MobileSAMv2/ultralytics/nn/autobackend.py), where the imgsz, names, or kpt_shape metadata fields are passed into a dynamically evaluated expression without neutralization (CWE-95). Publicly available exploit code exists, and the project has not responded to the issue report; exploitation requires local access and victim interaction to load the malicious model, limiting impact to Low confidentiality, integrity, and availability.