Scholarlink

In recent years, Side-Channel Attack (SCA) has been widely used in hardware security analysis. Such an attack uses the power characteristics generated by the switching of internal logic gates during chip operation to extract the waveform characteristics of the encryption key. Analyzing the complete encryption key statistically, it poses a great threat to the security of the chip. However, these attacks can be prevented by adding noise to the signal output pins of the chip. These noises can be divided into naturally generated Gaussian noise and artificial noise. Artificial noise can effectively defend against SCA especially. In order to improve the attack success rate, we aim to discuss the use of Autoencoder to reduce the interference of Random delay noise on SCA. In the experiment, we used Denoising Autoencoder (DAE) to preprocess the power traces generated during chip operation. Then we applied Correlation Power Analysis (CPA) to resolve the encryption key. In addition, we implemented the DAE processing mechanism in the ChipWhisperer-Nano experimental board and used the actual measured original power traces and noised power consumption traces for training. After completing the training model, it can effectively preprocess the noised data and key analysis. According to the experimental results, we have successfully improved the efficiency and success rate of analyzing keys.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Implementation and Analysis of Using Autoencoder to Suppress Noise Interference on Side-Channel Attacks

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Kuo-Yu Tsai,
  • Chung-Wei Kuo,
  • Wei-Ming Weng,
  • Guan-Ling Wang,
  • Chun-Chang Lin

摘要

In recent years, Side-Channel Attack (SCA) has been widely used in hardware security analysis. Such an attack uses the power characteristics generated by the switching of internal logic gates during chip operation to extract the waveform characteristics of the encryption key. Analyzing the complete encryption key statistically, it poses a great threat to the security of the chip. However, these attacks can be prevented by adding noise to the signal output pins of the chip. These noises can be divided into naturally generated Gaussian noise and artificial noise. Artificial noise can effectively defend against SCA especially. In order to improve the attack success rate, we aim to discuss the use of Autoencoder to reduce the interference of Random delay noise on SCA. In the experiment, we used Denoising Autoencoder (DAE) to preprocess the power traces generated during chip operation. Then we applied Correlation Power Analysis (CPA) to resolve the encryption key. In addition, we implemented the DAE processing mechanism in the ChipWhisperer-Nano experimental board and used the actual measured original power traces and noised power consumption traces for training. After completing the training model, it can effectively preprocess the noised data and key analysis. According to the experimental results, we have successfully improved the efficiency and success rate of analyzing keys.