STARM · HW-01 · HARDWARE
Hardware Backdoors
STARMCatalog entries from the DIPS Threat Inventory. The paper cites that dataset and does not use this name.
Malicious circuits implanted during chip manufacturing.
Severity in the dataset9/10
Not the paper’s H/M/L.
- Target
- Instructions
- Layer in the inventory
- ASICs (Application-Specific Integrated Circuits) / FPGAs (Field-Programmable Gate Arrays)
- Mitigation
- Secure chip design & visual inspection.
- How the inventory says to fix it
- AI/ML: Computer Vision for visual inspection of silicon dies to detect non-standard circuitry patterns.
- Quick fix
- N/A
- ML approaches named
- Random Forest (RF), Deep Neural Networks (DNNs) & Convolutional Neural Networks (CNNs), K-Nearest Neighbors (KNN), Support Vector Machine (SVM)
- Methodology
- By analyzing side-channel data (power, electromagnetic radiation, or timing) to identify anomalies that deviate from a "golden" (trusted) IC model
- Handler role
- Hardware Engineer
- Stage
- Manufacturing
- Standard named
- ISO 20243
- Status in the inventory
- Partially managed
STARMCatalog entries from the DIPS Threat Inventory. The paper cites that dataset and does not use this name.
Connection recorded in the inventory
Supply Chain Interdiction