AML.CS0013
Backdoor Attack on Deep Learning Models in Mobile Apps
- Type
- Exercise
- When
- 2021-01-18
- Actor, as ATLAS names it
- Yuanchun Li, Jiayi Hua, Haoyu Wang, Chunyang Chen, Yunxin Liu
- Target
- ML-based Android Apps
Techniques, in order
- 1Search Application RepositoriesReconnaissance · AML.T0004
- 2Acquire Public AI Artifacts: ModelsResource Development · AML.T0002.001
- 3Full AI Model AccessAI Model Access · AML.T0044
- 4Develop Capabilities: Adversarial AI AttacksResource Development · AML.T0017.000
- 5Manipulate AI Model: Modify AI Model ArchitecturePersistence · AML.T0018.001
- 6Verify AttackAI Attack Adaptation · AML.T0042
- 7Supply Chain Compromise: Model RepositoryInitial Access · AML.T0010.003
- 8Craft Adversarial Data: Insert Backdoor TriggerAI Attack Adaptation · AML.T0043.004
- 9Physical Environment AccessAI Model Access · AML.T0041
- 10Evade AI ModelImpact · AML.T0015
How these attacks work
Software used
- DeepPayload · Tool
From MITRE ATLAS 2026.10 (Apache-2.0). ShieldMap lists who and what was involved and the techniques used, and links to ATLAS for the write-up; it does not copy the steps. Attribution is ATLAS's, not ShieldPi's.