Research On Hardware Trojan Detection Method Based On Ring Oscillator

With the continuous development of electronic technology,various types of hardware and software have been applied to various industries,but the Trojan that appear in the application process seriously threaten the security of information management,and the hardware Trojan that exist in the chip are in danger.It is significantly higher than the software Trojan,which has aroused the attention of academia.First,this paper uses the EP4CE30F23C8 model FPGA to implement the SM3 encryption algorithm to simulate a dedicated encryption chip.There are 12 ring oscillators(Ring Oscillator,RO)set in the FPGA.Theoretically,the count of the ring oscillator with a hardware Trojan should be different from the count without a hardware Trojan.By observing the count,you can know whether there is a Trojan in the chip.The number of oscillations of the ring oscillator has a certain relationship with the temperature,so this paper conducts experiments under the environment of 25 ? 30 ?,35 ?,40 ? and 45 ?,and obtains the data.Secondly,if the obtained count is simply an average,the comparison effect is not ideal,so this paper uses the difference normalization algorithm to process the data,and the result graph after the difference normalization processing will be observed It was found that the difference between the ring oscillators implanted with hardware Trojans at room temperature is significantly different.Finally,using three machine learning algorithms to further process the data,through the obtained accuracy rate and the result graph,we can find that the accuracy rate obtained by using the logistic regression algorithm is above 94.67%;the accuracy rate obtained by using the K-means algorithm is at More than 98.5%;the correct rate obtained by using the improved support vector machine algorithm is also more than80%.This proves that the accuracy of using ring oscillators to detect hardware Trojans is very high and is a very efficient method.