Defense Mechanism Vulnerability Analysis of Ring Oscillator PUFs Against Neural Network Modeling Attacks using the Dragonfly Algorithm

Physical Unclonable Functions (PUFs) are used to provide security and authentication in assured and trusted integrated circuits by producing unclonable cryptographic keys. However, machine learning algorithm attacks have proven to be effective against PUFs. By using machine learning attacks the Challenge-Response Pairs (CRPs) can be predicted using a relatively small subset of training samples. In this work, Feed-Forward Neural Network (FNN) based model using the Dragonfly Optimization (DFO) is used to predict the accuracy of the response. The vulnerability analysis is performed on two well-known Ring oscillator PUFs; namely, the Configurable Ring Oscillator PUF and the XOR-Inverter Ring Oscillator PUF to analyze their defense mechanisms. From the study, it is found that the FNN-based model can predict the responses of the two PUFs with prediction accuracies of 85.2% and 71.3% respectively, thus making them vulnerable to this kind of attacks.