当前位置: X-MOL 学术Integration › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Improving power analysis attack resistance using intrinsic noise in 3D ICs
Integration ( IF 2.2 ) Pub Date : 2020-03-21 , DOI: 10.1016/j.vlsi.2020.02.007
Zhiming Zhang , Jaya Dofe , Qiaoyan Yu

Three-dimensional (3D) integration is envisioned as a natural defense to thwart side-channel analysis (SCA) attacks on the hardware implementation of cryptographic algorithms. However, neither physical experiments nor quantitative analysis is available in existing works to study the impact of power distribution network (PDN) on the SCA attacks. Through quantitative analyses and experiments with realistic 3D models, this work demonstrates the impact of noise in PDN on the 3D chip's resilience against correlation power analysis (CPA) attack, which is one of SCA attacks. The characteristic of PDN noise is extracted from our experiments. To expand the natural defense originated from the 3D integration, this work proposes to exploit the PDN noise inherently existing in 3D chips to thwart CPA attacks. Instead of introducing external noise or flattening the power profile, the proposed method utilizes the spatially and temporally varied supply voltages from other 3D planes to blur the power correlation of the crypto unit. Both theoretical analysis and experimental validation prove that the proposed method can effectively enhance the resilience of a crypto unit embedded in the 3D chip against CPA attacks. Simulation results show the proposed method improves the average guessing entropy by 9× over the baseline. Emulation on an FPGA platform demonstrates that the proposed method successfully slows down the key retrieval speed of CPA attack, with significantly less power overhead than representable power equalization techniques. Test vector leakage assessment (TVLA) shows that the proposed method improves the confidence to accept null hypothesis 201× over the baseline.



中文翻译:

利用3D IC中的固有噪声提高功率分析的抗攻击性

构想了三维(3D)集成,可以自然地防御对加密算法的硬件实现的边信道分析(SCA)攻击。但是,现有的研究既没有物理实验也没有定量分析来研究配电网(PDN)对SCA攻击的影响。通过定量分析和使用现实3D模型进行的实验,这项工作证明了PDN中的噪声对3D芯片抵御关联功率分析(CPA)攻击(这是SCA攻击之一)的弹性的影响。从我们的实验中提取了PDN噪声的特征。为了扩展源自3D集成的自然防御,这项工作建议利用3D芯片中固有的PDN噪声来阻止CPA攻击。代替引入外部噪声或使功率分布平坦化,所提出的方法利用来自其他3D平面的时空变化的电源电压来模糊密码单元的功率相关性。理论分析和实验验证均证明,该方法可以有效地增强3D芯片中嵌入的加密单元对CPA攻击的恢复能力。仿真结果表明,该方法将平均猜测熵提高了基线的9倍。在FPGA平台上的仿真表明,该方法成功降低了CPA攻击的密钥检索速度,与可表示的功率均衡技术相比,其功率开销明显更少。

更新日期:2020-03-21
down
wechat
bug