Profiled Power Analysis Attacks Using Convolutional Neural Networks with Domain Knowledge
Evaluation of cryptographic implementations against profiled side-channel attacks plays a fundamental role in security testing nowadays. Recently, deep neural networks and especially Convolutional Neural Networks have been introduced as a new tool for that purpose. Although having several practical advantages over common Gaussian templates such as intrinsic feature extraction, the deep-learning-based profiling techniques proposed in literature still require a suitable leakage model for the implementation under test. Since this is a crucial task, we are introducing domain knowledge to exploit the full power of approximating very complex functions with neural networks. By doing so, we are able to attack the secret key directly without any assumption about the leakage behavior. Our experiments confirmed that our method is much more efficient than state-of-the-art profiling approaches when targeting an unprotected hardware and a protected software implementation of the AES.
genetic algorithm data mining big datum power consumption data structure association rule data stream programmable gate array field programmable gate elliptic curve data mining technique efficient algorithm smart card fpga implementation association rule mining mining algorithm power analysi frequent itemset hyperspectral datum sliding window frequent pattern leaf area apriori algorithm mining association rule leaf area index side channel uncertain datum differentially private leakage power algorithmic approach mining association elliptic curve cryptosystem mining frequent itemset mining curve cryptosystem frequent itemset mining plant leaf power analysis attack differential power analysi item set data mining task data stream mining frequent item analysis attack differential power high utility stream mining mining frequent itemset chlorophyll content maximal frequent mining frequent pattern false negative data mining problem high utility itemset frequent closed frequent itemsets mining utility itemset association mining closed itemset chlorophyll fluorescence itemsets mining transactional datum efficient mining correlation power analysi side channel analysi dpa attack maximal frequent itemset frequent closed itemset mining problem mining maximal frequent mining data stream closed frequent mining maximal itemset mining algorithm simple power analysi mining frequent closed leaf chlorophyll content memory consumption leaf chlorophyll finding frequent closed frequent itemset maximum frequent discovering frequent koblitz curve weighted frequent mining closed estimating leaf vegetative growth cryptographic circuit fast mining airborne spectrographic imager chlorophyll meter compact airborne spectrographic finding frequent itemset mining closed frequent top-k frequent estimation of leaf leakage power analysi discovering frequent itemset transactional data stream parallel frequent weighted frequent itemset prosail model discovery of association approximate frequent mining top-k frequent parallel frequent itemset itemset mining problem probabilistic frequent itemset number of transactions frequent itemsets algorithm find frequent itemset