An efficient countermeasure against correlation power-analysis attacks with randomized montgomery operations for DF-ECC processor

  • Authors:
  • Jen-Wei Lee;Szu-Chi Chung;Hsie-Chia Chang;Chen-Yi Lee

  • Affiliations:
  • Department of Electronics Engineering and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan;Department of Electronics Engineering and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan;Department of Electronics Engineering and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan;Department of Electronics Engineering and Institute of Electronics, National Chiao Tung University, Hsinchu, Taiwan

  • Venue:
  • CHES'12 Proceedings of the 14th international conference on Cryptographic Hardware and Embedded Systems
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Correlation power-analysis (CPA) attacks are a serious threat for cryptographic device because the key can be disclosed from data-dependent power consumption. Hiding power consumption of encryption circuit can increase the security against CPA attacks, but it results in a large overhead for cost, speed, and energy dissipation. Masking processed data such as randomized scalar or primary base point on elliptic curve is another approach to prevent CPA attacks. However, these methods requiring pre-computed data are not suitable for hardware implementation of real-time applications. In this paper, a new CPA countermeasure performing all field operations in a randomized Montgomery domain is proposed to eliminate the correlation between target and reference power traces. After implemented in 90-nm CMOS process, our protected 521-bit dual-field elliptic curve cryptographic (DF-ECC) processor can perform one elliptic curve scalar multiplication (ECSM) in 4.57ms over GF(p521) and 2.77ms over GF(2409) with 3.6% area and 3.8% power overhead. Experiments from an FPGA evaluation board demonstrate that the private key of unprotected device will be revealed within 103 power traces, whereas the same attacks on our proposal cannot successfully extract the key value even after 106 measurements.