Table 1 Overview of HT countermeasures
From: Secure and Trustworthy Microelectronics: Vulnerabilities, Solutions, and Trends
Approaches | Reference | Detection/tolerance | Method | Weakness | Class of countermeasure |
|---|---|---|---|---|---|
Software variant | TReC [25] | Runtime | Execution of multiple functionally equivalent software variants | Performance overhead and not applicable for all kind of HT models | Trojan tolerance |
IC redundancy | SAFER PATH [26] | Runtime | Program execution in multiple processing elements | Performance and cost overhead with limited detection scope | Trojan tolerance |
IP redundancy | Runtime | IP vendor diversity | Performance and cost overhead with limited detection scope | Trojan tolerance | |
Static analysis | Pre-silicon | RTL code analysis using metrics line, statement, fsm, and toggle coverage etc. | Full coverage doesn’t guarantee HT-free design. Limited to designs combinational logic | Trojan detection | |
Formal methods | Pre-silicon | Convert the design to proof checking format and checks security properties | Detection scope is limited and need sophisticated comprehensive properties | Trojan detection | |
IP monitoring | Runtime | Modification of hardware code to monitor execution | Performance and cost overhead with limited detection scope | Trojan detection | |
Side-channel | Post-silicon | Detecting anomalies in power consumption, electromagnetic emissions, and timing characteristics | Golden model required, sensitive to noise | Trojan detection | |
Logic testing | Pre/post-silicon | Useful test-pattern generation and Boolean analysis to trigger rare logic conditions at internal nodes | Large HTs, test vector generation for large designs | Trojan detection | |
Design-for-Sec | Pre/post-silicon | On-chip trust evaluation modules for monitoring/detecting trojans in post-deployment stage | Performance and cost overhead with limited detection scope | Trojan detection/ prevention | |
Machine learning | Reverse engineering [34, 35], circuit feature analysis [36,37,38], side-channel analysis [39,40,41,42] | Pre/post-silicon | By utilizing different machine learning algorithms to use circuit features, i.e., structure, functionalities and parameters, the model tries to reveal the HTs or enhance the strategies | For rare and diverse trojan, it is challenging to build vast label data. Without golden reference, models struggle to generalize sophisticated HTs | Trojan detection |
Graph matching | Pre/post-silicon | A library consists of HTs graph and search those in the graph of 3PIP using graph matching algorithms | Only work for known HTs graph | Trojan detection | |
Hardware obfuscation | Pre-silicon | Increasing the complexity and obscuring the design and functionality | Increase complexity and design overhead with reliance on certain threat models | Trojan prevention | |
Split manufacturing | Pre-silicon | Cost and product deployment overhead | Dependency on external trust | Trojan prevention |