Browse / Model Evasion
Model Evasion
Inputs crafted so a model misclassifies or fails to detect. 17 records. Reference: ATLAS AML.T0015.
- Evade AI Model 4
- Craft Adversarial Data: Manual Modification 3
- Craft Adversarial Data: Black-Box Transfer 3
- Other weaknesses 2
- Craft Adversarial Data: Black-Box Optimization 2
- Craft Adversarial Data: White-Box Optimization 2
- Craft Adversarial Data: Insert Backdoor Trigger 1
- vLLM: Processing differential in multi-channel audio downmixing enables hidden-input/moderation bypass for audio models Content hidden in unused audio channels is heard by the model after mono downmixing but not by human listeners, evading moderation.
- vLLM: image EXIF Rotation & PNG tRNS Transparency Not Normalized, Causing Mismatch Between Model Input and Expectations Images with EXIF rotation or PNG transparency tricks are preprocessed differently than humans see them, so the model receives hidden or altered content.
- ProKYC: Deepfake Tool for Account Fraud Attacks
- Live Deepfake Image Injection to Evade Mobile KYC Verification
- Attempted Evasion of ML Phishing Webpage Detection System
- Confusing Antimalware Neural Networks
- Backdoor Attack on Deep Learning Models in Mobile Apps
- Bypassing ID.me Identity Verification
- Attack on Machine Translation Services
- Microsoft Edge AI Evasion
- Evasion of Deep Learning Detector for Malware C&C Traffic
- Botnet Domain Generation Algorithm (DGA) Detection Evasion
- Camera Hijack Attack on Facial Recognition System
- Microsoft Azure Service Disruption
- Face Identification System Evasion via Physical Countermeasures
- ProofPoint Evasion
- Bypassing Cylance's AI Malware Detection