Research Of Hardware Trojans Detection Technology Based On Side Channel Analysis

Hardware Trojans is the malicious circuits implanted in the IC design ormanufacturing process which makes integrated circuits fail under certain conditionsor disclose confidential information from the chip. Because of the hidden dangersand hazards of modern information systems due to hardware Trojans, Trojansdetection technology has become a research hotspot of information security areas.A hardware Trojans detection platform and detection flow based on sidechannel analysis have been designed in this paper. The Trojans circuits and carriercircuits is designed and verified based on power signal analysis. The applicationproblems in the ASIC chip detection is discussed and in-depth studies are carried outin the following areas:First, for the problem of Trojans activation in the detection, the circuit transitionprobability is modeled and the main factors of Trojans activation probability is analyzed. Aspecific module for Trojans Detection is proposed, the structure design, the key parameterselection and the insertion flow have been discussed in detail. The simulation results showthat this structure can raise transition probabilities of the all internal nodes to the setthreshold value with ten percent increase of area and shorten the testing time. To thedSFF-based structure, this structure has saved over30percent of area with the same Pth.Secondly, After the power model design, the problems of process variationnoise and data redundancy in the Trojans detection are analyzed and a datadimension optimization algorithm based on principal component analysis ispresented. Monte Carlo analysis was used to simulate the process variation in theexperiments. The simulation results show that this algorithm reduces the influenceof process variation in the detection and obtains a over fifty percents compressionrates of data dimensionalities.Finally, the determine mechanisms of Trojans detection are researched. Aftergiving the basic determine method, the application problem of the classicscorrelation coefficient in detection is discussed. Two determine parameters whichbased on region overlap ratio and weight respectively are design for detection ofTrojans types with the different power signal characteristic. The statistical analysisof simulation datas showed the proposed determine parameters have good robustness and effective detection can be achieved for hardware Trojans of morethan two percent of the total chip area under the influence of five percent processvariation.