Low-cost Standard Signatures in Wireless Sensor Networks: A Case for Reviving Pre-computation Techniques?

Effective pre-computation techniques have been proposed almost 15 years ago for trimming the cost of modular exponentiations at the basis of several standard signature and key management schemes, such as the (Elliptic Curve) Digital Signature Algorithm or Diffie-Hellman key exchange. Despite their promises, the actual application of such techniques in the wireless sensor security arena has been apparently overlooked, and most of the research effort has rather focused on the identification of alternative lightweight constructions. However, modern sensor are equipped with relatively large flash memories which make memory consumption a less critical requirement, and emerging energy harvesting technologies provide occasional energy peaks which could be exploited for anticipating otherwise costly computational tasks. These trends push for a reconsideration of pre-computation techniques, which are explored in this paper as follows: (1) we further optimize prior pre-computation techniques by exploiting more recent results on Cayley graph expanders, (2) we implement an ECDSA scheme relying on pre-computations over two different wireless sensor node platforms (TelosB and MICA2), and (3) we experimentally assess the relevant performance and energy costs. In the traditional scenario of wireless sensor networks without energy harvesting, our prototype ECDSA implementation, despite still not fully optimized, outperforms prior work by almost 50%, and achieves an efficiency superior to NTRU signatures, natural candidates for low-power devices. Finally, (4) we quantitatively discuss ways to exploit harvested energy peaks to further improve efficiency.