連續體束縛態之光柵結構應用於物理不可仿製功能

Abstract

近年來，隨著AI演算法與邊緣運算硬體設備快速成長，數據儲存與架構安全性的重要性也相應提高，數據的保密性與防篡改的要求亦隨之增長，除了透過軟體達成資訊保密外，物理不可仿製功能(PUF)作為一種基於硬體的安全防偽技術，利用製程中不可避免的隨機變數，如空氣、濕度、溫度等，於晶片的結構上產生微細差異。這些差異極難複製，從而降低了逆向解構風險。PUF的獨特隨機性和不可複製性賦予晶片猶如其獨特的「指紋」。 本研究探討連續體束縛態(BIC)在PUF技術中的應用。內容分為三個部分，首先透過數值分析與有限元素模擬，其BIC的品質因子(Q factor)在理論上趨近於無限大的特性，並透過改變光柵的填充因子(Fill factor)將亮模態與暗模態重疊形成奇異點(EP)，當結構尺寸發生細微的變化，模態的特徵頻率便會發生顯著改變。接著將模擬得到的反射頻譜，透過反傅立葉轉換至時域訊號，與輸入高斯雷射進行卷積，以模擬現實中身分辨識的挑戰響應對(CRP)過程。最後將接收到的訊號數位化成比特串，也就是晶片指紋，並且檢驗其隨機性、唯一性、可靠性等參數，並透過美國國家標準暨技術研究院的隨機性測試(NIST randomness test)作二次驗證，證明其可以被視為一個真正的隨機數產生器。 經過巨量數據統計分析(10k數量)，本研究所提出之BIC-PUF結構可被視為一個真正的隨機數產生器，其隨機性、唯一性、可靠性參數Ex、Ey、HDinter、HDintra，平均值分別為0.9595、0.9976、0.4969、0.0393；標準差分別為0.0986、0.0032、0.0351、0.0352，並且通過了NIST randomness test。本論文所提出之BIC模態PUF未來可以擴展至近紅外光與可見光應用。

In recent years, with the rapid growth of AI algorithms and edge computing hardware devices, the importance of data storage and architecture security has also increased, and the requirements for data confidentiality and anti-tampering have also increased. In addition to achieving information confidentiality through software, Physically Unclonable Function (PUF), as a hardware-based security technology, utilizes random variables that are unavoidable during the manufacturing process, such as air, humidity, temperature, etc., to produce minute differences in the structure of the chip. PUF, as a hardware-based security technology, utilizes unavoidable random variables in the manufacturing process, such as air, humidity, temperature, etc., to produce minute differences in the structure of the chip. These differences are extremely difficult to be replicated, thus reducing the risk of reverse deconstruction, and the unique randomness and non-replicability of PUFs give the chip a unique “fingerprint” as if it were a chip. This study investigates the application of Bound states In the Continuum (BIC) in PUF technology. The content is divided into three parts, firstly, through numerical analysis and finite element simulation, the quality factor (Q factor) of the BIC theoretically tends to the infinite characteristic, and by changing the fill factor of the grating (Fill factor), the light mode and the dark mode overlap to form the Exceptional Point (EP), and when the structural dimensions change slightly, the characteristics of the modes are frequently changed. When there is a small change in the structure size, the characteristic frequency of the mode will change significantly. The simulated reflection spectrum is then converted to a time-domain signal by inverse Fourier transformation and convolved with the input Gaussian laser to simulate the Challenge-Response Pair (CRP) process of real-world body recognition. Finally, the received signals are digitized into bit strings, i.e., chip fingerprints, and examined for parameters such as randomness (Ex、Ey), uniqueness (HDinter), and reliability (HDintra), and are verified as a true random number generator through the NIST randomness test. After a huge amount of statistical analysis (10k), the BIC-PUF structure proposed in this study has Ex、Ey、HDinter、HDintra, with mean values of 0.9595, 0.9976, 0.4969, 0.0393, and standard deviations of 0.0986, 0.0032, 0.0351, 0.0352, and has passed the NIST randomness test. The standard deviation is 0.0986, 0.0032, 0.0351, 0.0352, respectively, and has passed the NIST randomness test. The BIC mode PUF proposed in this paper can be extended to near-infrared and visible light applications in the future.