Improved Correlation On Power Analysis Attack Based On Preprocessde Power Traces

With the development of computer communication technology and electronic information technology, it becomes more and more important to protect the privacy of data and the security of communication. In our lives, the use of cryptographic equipment (such as IC card, SIM card, etc) is widespread, so ensuring the security of secret data in embedded devices and the security of network communication is become more and more challenging. When these devices work, some information such as power, electromagnetic will be revealled, which may be uesed by attackers. Side channel attacks (SCA, also known as bypass attack) as an important branch of cryptography, combined with the implementation of cryptographic algorithm and the leaked information can narrow the key space or directly get the key.In recent years, side-channel attacks have become a most powerful attack performed on cryptographic devices. And many side-channel attack methods have sprung up, such as time attacks, electromagnetic radiation attacks, power analysis attacks (including simple power attack, differential power attack, correlation power attack), fault attack, etc. Power analysis attack is the most simple and effective way of SCA, so it gets the high attention of researchers. And the correlation power attack has become the most common side-channel attack.In this paper, we introduce a method to improve the the correlation power attack. After collecting the power consumption of a cryptographic device when it encrypts or decrypts input data, we don’t directly analyses aligned power traces like CPA but firstly we preprocess the recorded power traces. In this step, we eliminate some noises in the original collected side-channel information..During the procedure, we introduce a four-dimension vector to express the basic unit which we deal with. We give the definition and some properties of this vector expression. And based the vector expression, we show that the preprocessed power traces(namely difference power traces) have less noises than the recorded power traces, so it increase the signal-to-noise ratio(SNR)comparing with non-preprocessed power traces. Later, we show that the difference Hamming weight model can simulate the difference power consumption well. After that we give the full steps of performing our improved CPA(named as ICPA). Then the experiment shows that the ICPA method has better performance and efficiency.