Related Work

Much of the work on encryption concentrates only on the design and security issues of the algorithm concerned. But in real systems, performance and implementation costs are as important as security. As described in the previous section, these issues are further amplified in a personal environment where some devices like Palm Pilot, for example, are inherently limited with respect to their processing ability, battery life etc., which in turn could severely affect the very feasibility of performing encryption. Unfortunately, there is very little work that analyses encryption performance on these small devices and networks. Our work fills this void and studies the feasibility of universal encryption considering the devices and networks in a personal environment.

The work that comes closest to ours is a study by Schneier et al, [Schneier99] which gives the relative performance of 15 different proposed submissions for AES (Advanced encryption standard) on 8-bit 32-bit and 64-bit machines. In a more recent study, Schneier and Whiting [Schneier00] analyze the performance of the 5 AES finalists on these machines in greater detail. These studies also give the memory requirements of the algorithms on 8-bit smart cards. But they omit power consumption costs which are important in a personal environment consisting of portable devices. J. Goodman and A. Chandrakasan [Goodman] explore low power encryption techniques for wireless networks. This work explores only hardware encryption techniques and suggests how best encryption can be optimized in hardware for low power consumption. Other work [Schneier97,Preneel,Touch95]study optimizations for various algorithms and how well encryption can be implemented in software.

All of the above work focus on the performance and costs with respect to one or two factors at the best. They do not comprehensively analyze all the attendant costs which are prohibitive for portable devices. Our work consolidates all earlier work and adds to those, to answer the question of universal encryption in a personal environment. We further study a few applications in greater detail to find the overhead that encryption adds with respect to latency and throughput.