Delay-Based PUF Racing Strategy for Entropy and Security
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional PUF systems face inefficiencies in generating maximal entropy with minimal PUF cells, leading to predictability issues and vulnerability to machine learning attacks, especially when using ring oscillators in multiple races.
Innovation Solution
Implementing a delay-based PUF device with a racing strategy that pairs winners against winners and losers against losers, limiting the number of rounds to the binary logarithm of the number of PUF cells to ensure full entropy while reducing the number of required PUF cells, and incorporating error correction to maintain consistency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional PUF systems use multiple races with ring oscillators, then entropy generation is improved, but predictability increases and security decreases
Solution Approach 1:
The patent implements dynamic pairing strategies where the pairing configuration changes based on the round number and previous race results. Specifically, the system uses different pairing modes (round-robin, random, or custom pairings) that are selected and updated dynamically during the racing process, preventing adversaries from predicting future pairings based on static patterns.
Solution Approach 2:
The system changes the parameter of pairing configuration across different rounds. The pairing strategy is adjusted based on the round number, using formulas such as pairing cell i with cell (i + round) mod N in early rounds, and transitioning to different pairing patterns in later rounds. This parameter变化 ensures that the same PUF cells are not repeatedly paired in predictable sequences.
2Device complexity
If the number of PUF cells is reduced, then device complexity is improved, but the ability to generate maximal entropy deteriorates
Solution Approach 1:
The patent enables each PUF cell to participate in multiple racing rounds continuously, extracting maximum entropy from each cell. By implementing multiple rounds of racing with dynamic pairings, the system ensures that each PUF cell is utilized repeatedly against different opponents, thereby generating N bits of entropy from N PUF cells across log2(N) rounds, achieving full entropy extraction without requiring additional cells.
Solution Approach 2:
The system employs periodic racing rounds with structured pairing patterns. Each round follows a specific pairing formula that systematically cycles through different cell combinations. For example, in round r, cell i is paired with cell (i + r) mod N, creating a periodic structure that ensures comprehensive entropy extraction while maintaining a manageable number of PUF cells.
3Loss of information
If multiple rounds of races are executed, then entropy extraction is improved, but error propagation increases
Solution Approach 1:
The patent incorporates feedback mechanisms where the results of each racing round are used to inform and adjust subsequent pairings and error correction strategies. The system monitors race outcomes and uses this feedback to refine pairing configurations in later rounds, ensuring that entropy is maximized while error patterns are identified and corrected. Error correction codes are applied based on feedback from previous round results.
Data Source
AI summary
According to various aspects, a delay-based physical unclonable function (PUF) device is provided. According to one embodiment, the PUF device includes circuitry for generating output bits of entropy by comparing, or “racing”, a plurality of PUF cells. A PUF cell is a building block of the PUF device. For example, the PUF device may include two identically designed circuits with only process related variations and each circuit can be a PUF cell. According to another aspect, if PUF cells with same history of winning or losing are being compared in a race, adversaries cannot predict the outcome of the current race based on previous race results. Accordingly, systems and methods are described herein for generating multiple rounds of races based on the previous rounds of races. Thus, one PUF cell can be used in multiple pairwise comparisons while maximal entropy is extracted.