As emerging sensor networks are normally deployed in the field and thus
vulnerable to many types of attacks, it is critical to implement cryptographic
algorithms in sensor nodes to provide security services. Public-key
algorithms, such as RSA and Elliptic Curve Cryptography (ECC), have been
widely used for security purposes like digital signature and authentication.
In resource-constrained sensor networks, ECC is more suitable than other
public-key algorithms based on large integer fields because it provides
similar security strength with shorter keys and thus is more
computation efficient. However, even ECC is slow on many sensor nodes. In
underwater sensor networks (UWSNs), sensor nodes communicate with each other
through acoustic channels. Because more complicated algorithms are needed to
encode and decode acoustic signals transmitted through the noisy underwater
channels, many acoustic modems include a DSP to meet the performance
requirement of underwater communications. In this paper, we study the
implementation of ECC on DSPs. We optimize the SECG elliptic curves secp160r1
and secp224r1 on a TMS320C6416 DSP board from Texas Instruments. In our
implementation, it takes 0.81 ms to compute a random scalar point
multiplication for secp160r1, an order of magnitude faster than the
communication algorithm decoding a data block. Therefore, we believe that it
is feasible to adopt ECC in UWSNs.