Signal processing for hardware trojan detection

Today, with increasing technological advances, trustworthiness has also been a major problem. Taking the production of IC as a global parameter is one of the strategies for reducing the chip cost. Advantages in manufacturing cost and an access to advanced fabrication facilities enforce manufacturer to outsource IC fabrication so that costs are reduced. Manufacturers could purchase specific parts of an electronic card, some ICs or embedded software designed by other person or companies on some ICs so that manufacturers utilize from knowledge of person or institution, which are specialist about specific topics. Software or hardware developed by these persons or institutions is called intellectual property. Person or institution that develop such software or hardware are called vendors. In all manufacturing stages of ICs, any adversary persons or institutions can interfere to process of design or production and introduce malicious software called hardware trojan through ICs. In this thesis, firstly some concepts with respect to manufacturing process, in which attackers subvert, are explained. Difference between trojans and viruses are discussed and features of trojans are briefly explained. Then methods of trojan detection are summarized and immediately after, signal processing method used for trajon detection is described a little detailed, since we will use this method to detect trojan signals. Summarizations of literatures about other trojan detections are presented. In Section 2, primarily it is explained why there is any visual difference between various transforms of trojan and clean signals. Theoretical background of signal processing techniques used in trojan detection, neural network and its performance are discussed. In Section 3, Trojan and trojan-free signals obtained from different pins of trojan IC and trojan-free IC are examined in time and frequency domain. Then trojan signals are detected by using signal processing techniques developed in Section 2. Conclusions section summarizes results of work.