Research On Electromagnetic Information Security Of Cryptographic Integrated Circuits Based On Side-channel Analysis

With the advent of the era of Internet of Everything(IoE),more and more intelligent hardware devices appear in modern life.At the same time,the problems of physical information security among intelligent hardware devices are gradually exposed.Cryptographic integrated circuit is the basis of intelligent hardware devices,and its security is often protected by cryptographic algorithm.Although the cryptographic algorithm design is secure in mathematical theory,it can not guarantee the security in physical implementation.Electromagnetic side-channel analysis is mainly a method to obtain the secret information by the electromagnetic side-channel leakage information in the process of implementing the cryptographic algorithm.At present,it has posed a serious threat to the physical information security of cryptographic integrated circuit.The advantages of electromagnetic side-channel analysis are non-invasive,strong concealment,easy to realize and has good analysis result,which has become one of the main side-channel analysis methods.In recent years,electromagnetic side-channel analysis technology has become the focus of research at home and abroad,especially in the field of side-channel analysis and detection of cryptographic integrated circuits,which has become one of the most promising directions of cryptanalysis.With the development of cryptographic integrated circuits,the current problems existing in electromagnetic side-channel analysis also include the following aspects:(1)There is no general model to explain the relationship between electromagnetic side-channel leakage of cryptographic integrated circuits and cryptographic algorithms,which is mostly based on the premise of the specific model.(2)Traditional electromagnetic side-channel analysis often uses a single byte of the key for analysis.In the process of hardware parallel implementation,there will be a waste of information,resulting in low signal-to-noise ratio and low analysis efficiency.(3)Hardware Trojan of cryptographic integrated circuit has the characteristics of diverse forms,strong concealment,low trigger probability and the difference in physical parameters caused by process deviations on cryptographic integrated circuits.Both of them make the detection of hardware Trojans difficult.The dissertation mainly focuses on the electromagnetic side-channel analysis and detection technology of cryptographic integrated circuits,aiming to solve the current problems existing in electromagnetic sidechannel analysis.Based on the Advanced Encryption Standard(AES)of block cipher algorithm,the dissertation analyzes the problems of the electromagnetic analysis model of information leakage in cryptographic integrated circuits,insufficient information utilization of traditional electromagnetic analysis in hardware parallel implementation and the difficulty of hardware Trojan detection.The main contents and contributions of the dissertation are as follows:(1)Electromagnetic analysis model of information leakage in cryptographic integrated circuitsIn view of the lack of a general model to explain the relationship between electromagnetic side-channel leakage of cryptographic integrated circuit and cryptographic algorithm,the working mechanism of cryptographic algorithm in cryptographic integrated circuit is studied.According to the current fluctuation caused by the key data change,the electromagnetic emission model of cryptographic integrated circuit is established by electric dipoles.At the same time,the near-field probe used to receive electromagnetic side-channel leakage is also analyzed,and the relationship between the output and the input is established according to its equivalent circuit to construct the electromagnetic receiving model.The electromagnetic emission model is the antecedent and the electromagnetic receiving model is the consequence.They together constitute the electromagnetic analysis model of information leakage in cryptographic integrated circuits.Finally,an electromagnetic analysis model based on multiple linear regression is constructed.The correctness of the proposed electromagnetic analysis model of information leakage is verified by the multiple linear regression theory in statistics.The model can be extended to general cryptographic integrated circuit equipment for electromagnetic side-channel analysis.(2)Multi-byte electromagnetic analysis based on intelligent optimization algorithm.In order to solve the problem of insufficient information utilization in the hardware parallel implementation of traditional electromagnetic sidechannel analysis,a multi-byte electromagnetic analysis method is proposed to improve the utilization of information.Particle swarm optimization algorithm and genetic algorithm are introduced to transform the key recovery problem of electromagnetic analysis into key optimization problem.Using the characteristics of iterative optimization and heuristic way of intelligent algorithm,the correct key can be found efficiently in the key search space.Through the improvement of intelligent algorithm,the key solution of cryptographic algorithm is more efficient.For particle swarm optimization algorithm,the inertia weight is added to the speed updating formula,while the mutation movement is added to the position update formula.The sorting operator and the sieve operator are added to the genetic algorithm.The improved intelligent algorithms not only improve the information utilization,but also greatly improve the key optimization.Compared with the traditional correlation electromagnetic analysis method,the multi-byte electromagnetic analysis method based on particle swarm optimization adopts a progressive method to achieve an improvement of about 13.72%in the number of electromagnetic emission traces.The multi-byte electromagnetic analysis based on genetic algorithm adopts a full coverage method to further reduces the traces,which has achieved an improvement of about 42.72%.(3)Hardware Trojan detection by electromagnetic side-channel based on transfer learning.Aiming at the problems of diverse forms,strong concealment,low trigger probability and the difference in physical parameters of hardware Trojan in cryptographic integrated circuit,a hardware Trojan detection framework based on transfer learning is proposed.Using the advantages that transfer learning can not only relax the limitation of training data,but also save training time,it has been applied to the detection of hardware Trojan and achieved good results.The time-frequency information of the collected electromagnetic side-channel signal of hardware Trojan is extracted by continuous wavelet transform,and the extracted timefrequency information is sent to the transfer learning network for detection and classification.The classification accuracy of 90.5%is obtained through the convolutional neural network,which is nearly 20%improved compared with the traditional support vector machine,and the accuracy of 90.9%is obtained in a relatively shorter time by transfer learning.In addition,transfer learning is used to extract the features of hardware Trojan and support vector machine is used as the classifier to replace the original classifier.Through the combination of transfer learning and support vector machine,the detection classification accuracy of hardware Trojan is further improved,which has achieved a classification accuracy of 94%.