| 摘要: | 在資訊安全時代之中，資料保護是很重要的一環，現在已經無法單純靠軟體來防護，在硬體安全防護方面，物理不可複製功能(Physical Unclonable Function, PUF)成為新的課題發展，靠著硬體本身製程上的變異性形成隨機字串而防止資料經由駭客盜取，在傳統揮發式記憶體的SRAM、DRAM已經逐漸達到物理極限，但另一方面，非揮發式記憶體較有發展空間，其主要面積大小、能耗和成本問題都能比揮發式記憶體中來的較佳，所以本篇論文探討利用電阻式記憶體為基礎設計一個1T1R陣列來實現以及驗證PUF功能，透過利用三種操作模式對電阻式記憶體的編程，使電阻式記憶體的阻態改變來區分記憶體儲存狀態0或1的訊號，並且利用這0與1訊號來進行PUF運用。
在實驗結果方面，可以看出在三種操作下以及兩種不同溫度下的情況來說，以SET操作方面較有各項的優勢，且在數據統計分析上，三種狀況下皆具有獨特性(Inter Hamming Distance)、可靠度(Intra Hamming Distance)和均勻性(Hamming Weight)等等，其在自相關係數(Auto correlation Function)也是都在95%區間之內，表示字串之間的相關性極低，本實驗三種操作狀態在25度常溫與75度高溫下皆通過美國國家標準與技術研究院(NIST)的15項字串隨機標準，並且進行分析使用XOR機制來進行PUF增強功能，增強安全性且能有效避免機器學習的建模攻擊以及側信道攻擊，因此本論文提出的設計利用大型陣列施行PUF功能，且能有效提高效率、增加安全強度係數與有效避免駭客攻擊。
;In the era of information security, data protection plays a crucial role, and it is no longer sufficient to rely on software for defense. Physical Unclonable Function (PUF) has emerged as a new frontier in the realm of hardware security. PUF leverages the inherent variability in the fabrication process of hardware to generate random strings, serving as a deterrent against data theft by hackers. Traditional volatile memory types, such as the SRAM and DRAM have gradually reached their physical limits. However, non-volatile memory (NVM) appears to offer more developmental potential, with advantages in terms of area size, power consumption, and cost, compared to volatile memory. Therefore, this paper explores the design and verification of a 1T1R array based on the RRAM to implement and validate PUF functionality.
Through three operational modes for programming resistive memory, the resistance state changes are utilized to differentiate between the stored signals representing 0 or 1 in the memory. The experiment results demonstrate that the SETSET operation exhibits superior performance in various aspects under three operational modes and two different temperature conditions. Statistical analysis shows that all three scenarios exhibit uniqueness (Inter hamming distance), reliability (Intra hamming distance), uniformity (Hamming weight), and low correlation between strings.
Furthermore, the proposed design passes NIST test at both room temperature (25 Celsius degrees) and high temperature (75 Celsius degrees). This paper introduces the use of the XOR mechanism to enhance the PUF function, thereby improving security and effectively mitigating machine learning modeling attacks and side-channel attacks. In conclusion, the designed PUF implementation utilizing a large-scale array proves to be efficient, enhances security strength, and effectively guards against hacker attacks. |
