Research On Hardware Trojan Detection Technology Based On Physical Characteristics

Science and technology growing,integrated circuit(IC),the core of modern electronic system,has been applied in many fields.While there is a research hotspot in the field of information security because of the particular IC industry recent years,that is the security of IC.As a new type of attack weapon,hardware Trojan refers to a redundant circuit that an "attacker" maliciously inserts without the circuit designer’s knowledge.Which can affect IC under certain conditions,for example,making IC less efficient,revealing key information and destroying system functions.IC designers expect to detect the hardware Trojan in each design stage to ensure a reliable IC design data due to the unreliability of IC design process.As an extra redundant circuit,hardware Trojan can change the physical characteristics of the circuit.So this paper carries out the following research as per the physical characteristics of IC such as structure,area and power consumption,machine learning algorithm as well as frequency domain analysis in order to solve the problem mentioned above.As for the detection of hardware Trojan in design stage,this paper gives a method based on machine learning for physical characteristics of IC to find the hardware Trojan,which is to say that the discriminant mechanism of hardware Trojan is equivalent to binary classification machine learning.The features of the training set and testing set adopted are mainly based on the physical information such as structure,area and power consumption showed in the original chip design stage and the chip design stage implanted with hardware Trojan.The AES encryption circuit Trojan library in Trust-Hub is selected as the reference circuit including 1 pure circuit and AES reference circuit implanted 21 different Trojans,and three machine learning algorithms including KNN,random forest and SVM are adopted to have tests in this paper.Results shows that this method can test all hardware Trojans implanted in the reference circuit,and the detection time for the selected reference circuit is less than 5ms.Further,it can layer the AES structure to locate hardware Trojans in the AES reference circuit on the basis of detecting them.Finally,this paper designs a hardware Trojan locating software based on AES reference circuit to make detection more efficient.One of the significant methods for detecting hardware Trojans in design stage is the hardware Trojan detection method based on side-channel signal analysis.The frequency domain analysis method is adopted in this paper to detect hardware Trojans as per the power dissipation of integrated circuit and hardware Trojans.The power dissipation information which can’t be discriminated in time domain is transformed to frequency domain by fast Fourier transform,the frequency domain information can be analyzed and sampled and then processed and judged by mahalanobis distance.It is proved that this method can detect all hardware Trojan circuits implanted in AES hardware Trojan library.