PUF Response Comparison for Hardware Trojan Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Physically unclonable functions (PUFs) are vulnerable to hardware trojan attacks and modeling attacks, which compromise their unique response generation and security, necessitating a method to detect and protect against these threats.
Innovation Solution
A system and method that generates and compares lists of PUF responses during production and in-field operation, monitoring for differences and patterns in challenge inputs to detect hardware trojans and modeling attacks, utilizing shared on-chip resources and varying supply voltage to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If PUF is used to provide unique signature and protect assets, then security and uniqueness are improved, but vulnerability to hardware trojan attacks and modeling attacks increases
Solution Approach 1:
The patent applies preliminary action by creating a reference list of PUF responses during production testing before the device is deployed. This reference list is stored and later used to compare against responses generated during field operation, enabling detection of hardware trojan attacks and modeling attacks through anomaly detection.
Solution Approach 2:
The patent implements feedback by continuously monitoring PUF responses during field operation and comparing them against the reference list. The system provides feedback through attack detection when discrepancies are found, enabling real-time security monitoring and response to potential attacks.
2Measurement precision
If list comparison method is used to detect attacks, then detection accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent applies segmentation by dividing the PUF response data into manageable lists that can be stored and processed efficiently. The reference list is created during production and stored in memory, while field responses are collected and compared in batches, making the complex comparison task more tractable.
Solution Approach 2:
The patent uses copying by creating a reference copy of the PUF response list during production testing. This reference copy is stored and used for comparison against field responses, eliminating the need to re-analyze original production data and reducing computational complexity during field operation.
3Reliability
If multiple challenges are provided to detect modeling attacks, then detection capability is improved, but time consumption and resource usage increase
Solution Approach 1:
The patent applies preliminary action by pre-collecting and storing multiple PUF responses during production testing in the reference list. This preparation enables efficient detection during field operation without requiring time-consuming real-time analysis of individual challenges.
Solution Approach 2:
The patent implements continuity of useful action by continuously collecting PUF responses during field operation and comparing them against the reference list. The system maintains ongoing monitoring without interruption, efficiently utilizing the pre-prepared reference data to detect modeling attacks over time.
Data Source
AI summary
A method and data processing system is provided for detecting an attack on a physically unclonable function (PUF). In the method, a first list of PUF responses to challenges is produced during production testing of an integrated circuit comprising the PUF. The first list is stored in a memory on the integrated circuit. A second list of PUF responses to the challenges is produced during normal operation of the integrated circuit in the field. The second list is compared to the first list. A difference between entries of the first and second lists computed. If the difference is greater than a threshold difference, then an indication of a hardware trojan is generated. The method may also include monitoring a series of challenges for an indication of a non-random pattern in the series. Detection of a non-random pattern may indicate a modeling attack.