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

Abstract

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