基于主成分分析的硬件木马检测技术研究

发布时间：2018-01-27 05:07

本文关键词： 硬件木马检测功耗预处理主成分分析距离判断　出处：《天津大学》2016年硕士论文　论文类型：学位论文

【摘要】：硬件木马是在集成电路设计或制造中对电路的恶意篡改,一旦激活工作,将窃取关键信息或者使芯片失效。针对硬件木马的检测技术得到广泛研究,其中基于侧信道信息分析的硬件木马检测是目前研究较多的检测方法。侧信道检测方法主要是通过分析侧信道信息中硬件木马引入的差异实现检测,而数据处理方法对实现有效的检测至关重要。本文利用侧信道分析技术,基于主成分分析法提出了一套数据处理识别算法并采用功耗分析的方法进行验证。首先对植入了硬件木马电路的功耗信息建立模型,并基于该模型设计了载体电路和硬件木马电路。同时优化测试环境和测试技术,从基于FPGA的硬件木马检测平台中获取了母本电路功耗和植入了硬件木马的电路功耗。然后,针对获取的功耗数据样本,分析数据特点,采用相应的数据处理方法优化数据,分类识别。重点是采用预处理方法解决数据的波形未对齐、异常值、噪声问题,并基于主成分分析法实现了母本数据和含硬件木马数据的特征提取和特征选择,得到的主特征涵盖了原数据99%以上的信息,再采用距离判断的方法实现硬件木马的有效检测。最后,针对基于主成分分析的硬件木马检测方法,优化预处理算法,提高检测精度。同时,在改变母本样本量、采样频率等参数的情况下,验证本方法的适用性,并基于MATLAB开发了硬件木马数据处理系统。结果表明,基于主成分分析结合距离判断的方法能够有效检测出占母本电路面积为0.15%左右的硬件木马。
[Abstract]:Hardware Trojan is the malicious tampering of the circuit in the design or manufacture of integrated circuits. Once activated, it will steal critical information or invalidate the chip. The detection technology of hardware Trojan has been widely studied. The hardware Trojan detection based on the side channel information analysis is the most widely studied detection method. The side channel detection method is mainly through the analysis of the differences in the side channel information introduced by the hardware Trojan horse to achieve detection. The data processing method is very important to realize the effective detection. This paper uses the side channel analysis technology. Based on principal component analysis, a set of data processing identification algorithm is proposed and validated by power analysis. Firstly, the model of power consumption information of Trojan circuit is built. Based on the model, the carrier circuit and the hardware Trojan circuit are designed. At the same time, the test environment and test technology are optimized. From the hardware Trojan detection platform based on FPGA, the power consumption of the mother circuit and the circuit power of the implanted hardware Trojan are obtained. Then, the characteristics of the data are analyzed according to the obtained power consumption data sample. Using the corresponding data processing method to optimize the data, classification and recognition. The emphasis is to use the pre-processing method to solve the data waveform unaligned, abnormal values, noise problems. Based on the principal component analysis method, the feature extraction and feature selection of mother data and Trojan horse data are realized. The main features cover the information of the original data more than 99%. Finally, aiming at the hardware Trojan detection method based on principal component analysis, the preprocessing algorithm is optimized to improve the detection accuracy. The applicability of this method is verified by changing the sample size and sampling frequency, and the hardware Trojan data processing system is developed based on MATLAB. The method based on principal component analysis (PCA) combined with distance judgment can effectively detect the Trojan horse which occupies about 0.15% of the female circuit area.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN407;TP309

【相似文献】