【作者】 程鹏；

【导师】 白国强；

【作者基本信息】 清华大学 ， 集成电路工程， 2015， 硕士

【摘要】 在当前的信息化时代,基于集成电路技术的智能IC卡或各类手提式设备作为各种信息系统用户的终端设备已被普遍使用。在这些用户终端设备中一般包含一个或多个基于集成电路技术的非易失性存储器用以保存用户的个人敏感信息(包括用户个人身份信息、用于身份识别和通讯保密的密钥等)和系统的核心参数(例如数字电视系统中的根密钥)。如何保护这些敏感信息不被非法获取是保护用户权益,保障相应信息系统正常运行的基本要求。本文尝试使用半侵入式与侵入式攻击技术对半导体非易失性存储器内容进行提取,以便研究半侵入式侵入式攻击方法和检测半导体非易失性存储器在半侵入式和侵入式攻击下的抗攻击能力。论文主要介绍基于光学方法和探针方法对存储器内容尝试提取的研究成果。本文首先使用激光错误注入方法尝试对智能卡芯片在验证过程中的验证数据进行提取,使用Riscue设备可以精确控制错误注入的时间和错误值,从而减少穷举猜测验证数据的次数。由于从正面进行攻击受到多层金属层阻挡,因此没有成功注入错误,需要从芯片背面加以尝试,此方法实现精确控制错误。其次,本文采用光化学方法对一个通用EEPROM存储器芯片直接使用扫描电镜(SEM)进行观测,尝试从图像上辨别不同存储状态的单元并且定位单个存储器单元位置。结果显示虽然无法直接依据图像分辨存储内容,但是可以清晰定位每个存储器晶体管的具体位置。再次,本文针对存储器整体的探针攻击直接探测EEPROM数据输出口数据,通过探针探测焊盘(PAD)信号从而读取存储器内容,实验显示这种攻击的可能性,但是工程难度大,此方法的创新点和意义在于攻击点新颖,成功处理芯片表面Dummy金属;最后,本文使用探针探测EEPROM每个单元的工作电信号,从而区分不同状态的单元,实验结果显示从I-V曲线上能明显区分存储器单元的不同状态,证实方法可行,此法攻击直接,攻击不需要了解目标芯片资料,实验成功消除探针噪声干扰,找到更好的制作焊盘(PAD)的方法。通过本文研究,发现半侵入式与侵入式攻击提取非易失性存储器内容实际操作工程难度大,但是攻击方式直接,可以弥补非侵入式攻击的不足,具有进一步研究的价值。此外,非易失性存储器对这两类攻击有一定抵抗能力。更多还原

【Abstract】 This paper searched how to use semi-invasive and invasive attack techniques to extract contents from nonvolatile semiconductor memories(NVSM). As sensitive information such as keys of encryption algorithms and passwords which needs to be kept for long time are stored in this kind of memories, storage security of NVSM is the core part of chip security. Only after grasping attack methods can effective defensive strategies be designed. Therefore, this paper explored which attack techniques can be used to extract contents from NVSM.There are three classes of attack technologies. They are non-invasive attack, semiinvasive attack and invasive attack. In this paper, several semi-invasive and invasive attack methods would be used to extract contents from NVSM.This paper proposed four attack methods, showed their process and results. Four methods are optical fault injection, light chemistry invasive attack, probing attack aiming at memory block, and probing attack aiming at a single memory cell. The first method used laser to inject fault when a smartcard executed codes to guess contents more easily with brute force attack. It is hard to inject laser from front side because of multiple metal layers of the target chip, and injecting form back side should be tried. The creative point of this method is trying to inject faults at exact moment to control the faults. The second experiment used scanning electron microscope to observe a preprocessed general EEPROM memory chip to distinguish and locate different storage state cells. Although the result showed contents could not be read directly but every single transistor of memory cells could be located. The creative part and meaning of this experiment is combining analysis ways of failure analysis and data security, and the result can be used to help with reducing time cost of reverse engineering. The third experiment used needles to test signals by placing needles on PAD which connected data output ports of EEPROM, and the process showed the possibility of this attack way, but this method is difficult to execute from engineering perspective. The creative point of this attack lies in its unique attack point and it successfully remove the passive influence of dummy metal. The last experiment used probing needles to test working signals of a single memory cell by contacting PADs which connected electrodes of memory cell transistors to distinguish different cells, and the result showed that different state of memory cells could be shown directly, which verified this attack could work. This attack method is direct and it can be launched even without any information of target chips. Meanwhile, this experiment successfully removed background noise which does harm to probing, and found better way to deposit PAD..Overall, using invasive and semi-invasive attack methods to extract contents from NVSM is difficult in engineering perspective, but this topic deserves further research because theses attack ways are direct, and they can cover the shortage of noninvasive attack.更多还原