Feasibility Evaluation of Neural Network Physical Unclonable Function
Kazuya Shibagaki, Taichi Umeda, Yusuke Nozaki, Masaya Yoshikawa · 2018
In recent years, consumer electronics that utilize artificial intelligence (AI) have attracted attention. As AI is widely used, interest in the security of AI has been heightened. We have proposed a concept of the neural network physical unclonable function (NNPUF) which was suitable for AI edge modules and it utilized neural network propagation delay time as authentication. This study applies the NNPUF to benchmark problems to evaluate its feasibility. Experimental results prove the feasibility.