1200μm² Physical Random-Number Generators Based on SiN MOSFET for Secure Smart-Card Application

Abstract

Secure digital chips such as those found in smart cards are widely used for financial transactions and the transfer of confidential information. Small circuits for high-level information security have to be implemented in these chips. These secure circuits thus require a small ph-RNG (physical random-number generator) capable of generating unpredictable random numbers. A smart card of just a few mm 2 is almost entirely occupied by the CPU, coprocessor, random logic, ROM, RAM, EEPROM, etc., and the memory capacity required is increasing. However, the circuit area of previous ph-RNGs is large, since there is a complex trade-off among circuit area, quality of random numbers and data generation rate. To solve the tradeoff, we use SiN MOSFET as a noise source device, and design a compact A/D converter for SiN MOS-FET RNG. Together these components form a compact ph-RNG circuit as small as 1200μm 2 with a generation rate of 2Mb/s. This ph-RNG circuit area is smaller than one-third of that of any ph-RNG previously reported [1][2][3][4][5], compared at the same generation rate. Furthermore, the quality of random numbers generated by the SiN MOSFET ph-RNG does not depend on temperature, since the origin of random noise signal is a direct tunneling process, theoretically independent of temperature, whereas the origin of random noise in previous ph-RNG is from thermal processes.Previous ph-RNGs are based on thermal noise or thermal carrier injection/ejection in a single or few local traps [1][2][3][4][5]. These noise signals are very small, especially at high frequency, although large noise signals at high frequency are better for a ph-RNG. We use a MOSFET with high-density electron traps in a SiN layer near a Si channel for a noise source ( Fig. 22.8.1). The SiN layer is formed by the same CVD used in the conventional CMOS mass-production process. The SiN MOSFET is fabricated with a CMOS process and one additional photo mask. Electrons are injected/ejected quickly between the Si channel and local traps in SiN layer by direct tunneling. According to electron injection/ejection in many local traps, a large conductance change in the transistor is seen as a large drain current (I d ) fluctuation in the SiN MOSFET. The noise signal of SiN MOSFET is much larger than that of a conventional MOSFET, as shown in Fig. 22.8.1. The circuit blocks for conventional ph-RNG and the SiN MOSFET one are shown in Fig. 22.8.2. Most conventional ph-RNGs require large amplification of the noise signal or a large array of identical noise generators and A/D converters in order to attain over 1Mb/s generation rate, since the noise signals are very small at high frequency. As a result, the amplifier or A/D converter has been very large. In this work, because of the high-amplitude random noise at high frequency from the SiN MOSFET, we need only a single amplifier and A/D converter, and the amplifier area is decreased, as shown in Fig. 22.8.2.In addition to an amplifier and A/D converter, a RNG also needs filters to cut off low-frequency noise origina...

High-Speed Magnetoresistive Random-Access Memory Random Number Generator Using Error-Correcting Code

“…Here, we investigate N-bit LFSR with N ≤ 8. Concretely, we try (1, 0), (2, 1, 0), (3, 1, 0), (4, 1, 0), (7,1,0), and (7, 3, 0) LFSRs, where listing (1, 0) means that a new bit is generated by XORing a shifted bit with the next bit, and listing (7, 1, 0) means that a seven times shifted bit and a shifted bit are XORed with the next bit, and so on. Figure 5 shows (2, 1, 0)-LFSR and (3, 1, 0 …”

High-Speed Magnetoresistive Random-Access Memory Random Number Generator Using Error-Correcting Code

“…Here, we investigate N-bit LFSR with N ≤ 8. Concretely, we try (1, 0), (2, 1, 0), (3, 1, 0), (4, 1, 0), (7,1,0), and (7, 3, 0) LFSRs, where listing (1, 0) means that a new bit is generated by XORing a shifted bit with the next bit, and listing (7, 1, 0) means that a seven times shifted bit and a shifted bit are XORed with the next bit, and so on. …”

“…Table III shows that the estimated generation speeds of MRAM-based RNGs are much faster than those of state-ofthe-art RNGs reported in ref. 1 , as a function of error correcting capability t for n = 31, 63 and 127 BCH codes. Note that raw bit sequence just after MRAM unit fails in almost all these tests.…”

“…RNGs capable of generating natural random numbers by means of physical phenomena have recently been investigated and have begun to replace pseudo-random numbers generated by software. Various kinds of on-chip RNG using physical noise signals such as random telegraph noise in transistors [1][2][3][4][5][6] , noise in diode junctions, or magnetic tunnel junctions (MTJs)for magnetoresistive random-access memory (MRAM) [7][8][9] have been proposed.…”

Générateur analogique chaotique intégré faible consommation

“…Rigaud (4) (1) ISEN Toulon et pole CNFM de Toulon (2) Polytech'Marseille et pole CNFM de Marseille (3) IM2NP Laboratoire, (4) EMSE, Ecole des Mines de Saint Etienne Contact: edith.kussener@yncrea.fr Dans le cadre de la formation à l'ISEN Toulon, les étudiants de première année de Master ont pour mission de réaliser un projet sur une durée de 300h. Ce projet a eu pour vocation première la réalisation d'un produit en utilisant les outils d'aide à la simulation proposés par le CNFM, mais également les outils de gestion de projets.…”