Deep learning attack for physical unclonable function
Yoshiya Ikezaki, Yusuke Nozaki, Masaya Yoshikawa · 2016
The semiconductor counterfeiting has become a serious problem. Several Physical Unclonable Functions (PUFs), which utilizes the variation when manufacturing, are proposed as a countermeasure for imitation electronics. An arbiter PUF is one of the most popular PUFs. The operation of an arbiter PUF can be expressed by using a delay model. An arbiter PUF is reported to be attacked by forcing them to learn the delay model. Almost all of previous studies used SVM for the learning. This study proposes a new attack method using a deep learning technique. Experiments prove the validity of the proposed method.