Runtime Detection of Malicious ML Models (ModelShield)

Evidence record

ModelShield scanner evaluation including difficult independent holdout cases.

Read the original source ↗

Claim-level citations (3)

• supportsModelShield detected malicious model execution but missed many independent holdout samples: The paper evaluates kernel and framework runtime observations to detect active malicious serialized model behaviors.

  Methods and attack behavior classification

• supportsModelShield detected malicious model execution but missed many independent holdout samples: Independent holdout results include 47 detected out of 84 malicious artifacts, illustrating dormant/conditional evasion.

  Evaluation, independent holdout and false negatives

• supportsModelShield detected malicious model execution but missed many independent holdout samples: On the independent 84-malicious, 252-benign evaluation set, ModelShield detected 47 malicious samples, missed 37, and produced no false positives, giving 55.95% recall.

  Section VI-A, Table IV: independent holdout detection results