The IoT industry faces many challenges, one of the most important of which is communication security. Also, one of the most critical actions for establishing security in communications is authentication; This means who sent the received message. Due to replay attack and other attacks in this area, the attacker can send the incomprehensible message received at the right time and achieve his sinister goals even if he achieves an encrypted message. Therefore, even the complexity of data encryption does not lead to network security. As a result, the use of cryptographic schemes alone can not be a good solution for authentication. The proposed method for authentication is to use the physical parameters. These parameters are randomly generated in the manufacturing process, so it is uncontrollable to make changes. These protocols are called Physical Unclonable Function. In this study, the resistance of some of these protocols has been investigated. Due to the advantages and disadvantages of these models, a new model called SQ-PUF has been introduced, which can show good resistance to attacks based on machine learning. This is a secure and scalable architecture. Indeed, we devised two alternative learning strategies to attack RPUF and OBPUF that are orders of magnitude more efficient.
Finally, we examined the resistance of this model against machine learning attacks and reports on the uniformity and uniqueness of data and hardware overhead of the above model.
Supervisor: Dr. Bijan Alizadeh
An Electronic copy of my M.Sc. thesis (Persian)
