PUF Authentication via Adversarial Challenge Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Physically Unclonable Functions (PUFs) are vulnerable to modeling attacks using machine learning or neural networks, allowing sophisticated attackers to clone devices even with previously unused challenges, despite their design to prevent replay attacks.
Innovation Solution
Implementing stability-based adversarial challenge selection techniques that utilize unstable Challenge-Response Pairs (CRPs) during device authentication, characterizing PUF devices across voltage and temperature ranges to disrupt attacker models and increase complexity, without requiring additional PUF circuit changes or degrading response reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If PUF CRPs are used only once to eliminate replay attacks, then replay attack resistance is improved, but sophisticated attackers can still monitor CRPs and use machine learning models to learn PUF behavior and clone devices
Solution Approach 1:
The system preemptively applies countermeasures by selecting challenges that are predicted to produce unstable responses based on environmental condition analysis. This preliminary action prevents attackers from obtaining reliable training data before the authentication occurs, countering modeling attacks in advance while maintaining replay attack resistance through one-time CRP usage
Solution Approach 2:
The system changes the stability parameter of challenge responses by selecting challenges whose response stability varies under different environmental conditions (temperature, voltage, frequency). This parameter change ensures that responses are unreliable for modeling purposes while remaining functional for authentication, resolving the contradiction between replay protection and modeling attack vulnerability
2Reliability
If stable challenge responses are used for authentication, then response reliability is improved, but attackers can more easily train machine learning models on consistent responses
Solution Approach 1:
The system makes the response stability dynamic by selecting challenges whose stability characteristics change based on environmental conditions. Challenges are chosen such that their responses are stable under current operating conditions for reliable authentication but would be unstable under varying conditions, making them unsuitable for model training. This dynamic approach maintains response reliability while increasing attack complexity
Solution Approach 2:
The system performs preliminary analysis of environmental conditions (temperature, voltage, frequency) to predict which challenges will produce unstable responses. This preliminary action allows the selection of appropriate challenges before authentication occurs, ensuring both reliable authentication responses and difficult-to-model behavior for attackers
3Reliability
If additional PUF circuit changes are implemented to improve security, then resistance to modeling attacks is improved, but device complexity and manufacturing cost increase
Solution Approach 1:
The system improves modeling attack resistance by changing operational parameters (selecting challenges based on predicted response stability under environmental conditions) rather than changing the physical PUF circuit structure. This approach maintains simplicity of the PUF device itself while achieving enhanced security through intelligent challenge selection, avoiding increased device complexity and manufacturing cost
Data Source
AI summary
A method comprises generating, during an enrollment process conducted in a controlled environment, a dark bit mask comprising a plurality of state information values derived from a plurality of entropy sources at a plurality of operating conditions for an electronic device, and using at least a portion of the plurality of state information values to generate a set of challenge-response pairs for use in an authentication process for the electronic device.