基于RO硬件木马检测的工艺偏差校正方法

Process Variation Calibration Method for RO-based Hardware Trojan Detection Technology

针对基于环形振荡器(ring oscillator,RO)的硬件木马检测方法受到工艺偏差的严重影响,导致木马检测准确度降低的问题,提出了一种新的针对RO硬件木马检测进行工艺偏差校正的方法.首先,根据硬件木马和工艺偏差对芯片供电电压变化的不同反应特性,改变测试数据中两者所占的比重,获取大致工艺偏差影响范围,然后削弱测试数据中工艺偏差的影响,突出和显化硬件木马的影响,最后通过马氏距离以及欧式距离等判别方法识别出硬件木马.利用现场可编程门阵列(field-programmable gate array,FPGA)测试平台进行了实验验证,数据处理结果表明,在0.61%,片内工艺偏差下可以有效检测出工艺偏差校正前无法识别的硬件木马,木马电路的大小等效为44个与非门.并且提出的方法可以扩展应用于包含片内工艺偏差的情况.

Ring oscillator（RO）based hardware Trojan（HT）detection technology is severely influenced by process variations,which leads to the decrease of the Trojan detection precision. Therefore,a new process variation calibra-tion method for RO-based HT detection technology is proposed in this paper. Firstly,according to the distinct re-sponse to the supply voltage of the HTs and process variations,respective proportion of the two in the test data is changed and approximate process variation amplitude is obtained. Secondly,the impact of process variation is cali-brated to lower its effect on the oscillation frequency and the impact of HT is manifested. Finally,pairwise distance analysis method using Mahalanobis distance or Euclidean distance is applied to identify whether the circuit is Trojaninfected. The proposed method is validated on the FPGA platform and the results show that within 0.61%, intra-die process variation,HTs that cannot be detected before the calibration can be effectively identified,and the size of the trojan circuit is equivalent to 44,NAND gates. Further,the proposed method can be extended to applications which include the impact of inter-die process variation.