Efficient implementation of bilinear pairing-based cryptosystems for sensors

Abstract

Since the introduction of bilinear pairing to public key cryptography in 2001, pairing has been considered as one of the most expensive public key operations in terms of both computational complexity and memory requirement. Recently some work has been done on improving the computation time of pairing on resource-constrained wireless sensors. However, little has been focused on neither reducing the memory consumption, nor providing a fully functional pairing-based cryptographic library for Wireless Sensor Networks. In this thesis, we propose three new algorithms for speeding up the computation and reducing the memory footprint of cubing, modular reduction and polynomial multiplication in ηT pairing over finite fields of characteristic three. We further propose new programming techniques for making the implementation even more lightweight. Experimental results show that one ηT pairing in GF(397) can be done on a MICAz sensor node in just 5.3 seconds, using only 154 bytes of RAM and 8,576 byes of ROM. As we know, a complete pairing-based cryptographic scheme requires much more than just a bilinear operation. Besides the work on pairing computation, we also present the first fully functional pairing-based cryptographic library for WSN. The library is fast and lightweight, and has an additional of one identity-based encryption scheme and two short signature schemes included. The performance results of implementing the three pairing-based cryptographic schemes show that pairing based cryptosystems are feasible and applicable in WSNs. In particular, the amount of RAM and ROM taken by each of these pairing-based cryptographic schemes is no more than 10% and 20%, respectively, of the total capacities of a MICAz mote.