Title Page
Abstract
Contents
Chapter 1. Introduction 13
1.1. Physical Unclonable Functions 13
1.2. PUF Performance Metrics 17
1.3. NAND Flash memory 21
1.4. Charge Trap Capacitor 23
1.5. Outline of the Dissertation 25
Chapter 2. Device Characteristics 26
2.1. Fabrication Process Flow 26
2.2. Measurement setup & Electrical Characteristics 31
Chapter 3. 2-D PUF Characteristics 35
3.1. Array Level Characteristics 35
3.2. CRP Generation Method 41
3.3. PUF Characteristics 46
Chapter 4. 3-D Memcapacitor Simulation 53
4.1. CRP Generation Methods 53
4.2. SMARTSPICE simulation 57
4.3. 3-D PUF Characteristics. 59
Chapter 5. Conclusion 64
Appendix A. Machine-learning attack Method 66
Bibliography 68
초록 74
Table. 3.1. NIST 800-22 randomness test results using 1M 1-bit responses based on experimental results. 52
Table. 4.1. NIST 800-22 randomness test results using 1M 1-bit responses for 3D array configurations. 63
Fig. 1.1. (a) User's data authentication process and Hacker's attack using machine learning model. (b) Properties of PUF. 16
Fig. 1.2. Four performance metrics of PUF (a) uniformity (b) diffuseness (c) uniqueness (d) bit error rate (reliability). 20
Fig. 1.3. (a) Array structure of NAND flash memory (b) Flash memory basic structure showing mechanism and transfer curve. 22
Fig. 1.4. Schematic of NAND flash array and capacitor crossbar array. 24
Fig. 1.5. NAND flash string structure according to perspective. (a) transistors (b) gate capacitors. 24
Fig. 2.1. Process flow and cross-sectional view of memcapacitor crossbar array. 28
Fig. 2.2. Array SEM image and cross-sectional TEM image of memcapacitor having TANOS stack. 29
Fig. 2.3. EDS result of fabricated memcapacitor having TANOS stack with charge trapping layer. 30
Fig. 2.4. Experimental setup for array-level characterization and PUF system. 33
Fig. 2.5. Electrical characteristics of TANOS capacitor obtained at single device level. (a) C-V graph (b) integrated charge. 34
Fig. 2.6. Pulse scheme method of memcapacitor crossbar array read operation. 34
Fig. 3.1. (a) OM image of 32X32 Memcapacitor crossbar array. (b) C-V characteristics of 1,024 devices in 32X32 crossbar array. (c) Charge... 38
Fig. 3.2. Read operations of memcapacitor crossbar array depending on the number of selected WLs (0~32). 39
Fig. 3.3. Accumulated charge during the read operation according to the number of the selected WLs with randomly generated 1,000 input signals in four arrays. 40
Fig. 3.4. (a)A PUF system with 32x 32 memcapacitor crossbar array. (b) Experimental setup for verifying PUF CRP extraction methods with sensing... 44
Fig. 3.5. (a) Voltage output through the integrator depending on the number of selected WLs. (b)Box plot of V out from different 32 BLs due to device variations. 44
Fig. 3.6. Measured comparator operations of generating response (a) '1' and (b) '0'. 45
Fig. 3.7. (a) Fractional Hamming weight distribution for uniformity. (b) Fractional Hamming distance distribution for diffuseness. (c) intra Hamming... 50
Fig. 3.8. (a) Speckle diagram of 1k 496-bit keys for four crossbar PUF chips. (b) Autocorrelation test results of security keys for PUF chips. (c) Machine... 51
Fig. 4.1. A schematic view of 3-D memcapacitor array structure based on 3-D NAND flash architecture. 55
Fig. 4.2. Two methods of extracting CRPs according to the use of a plane in a 3-D memcapacitor array. 56
Fig. 4.3. VMM operation results based on 16X16X16 3-D memcapacitor array in CRP and Multi-chip modes. 58
Fig. 4.4. (a) Hamming weight distributions for uniformity, (b) inter Hamming distance distributions for diffuseness, and (c) intra Hamming distance... 61
Fig. 4.5. Auto correlation results for (a) Multi-chip mode. (b) CRP mode. (c) Machine learning attack characteristics of two methods. (d) Training and test... 62
Fig. A.1. (a) Network structure used in PUF's machine learning attack. (b) Various parameters used in the network learning process. 67