Biomedical microelectronic implants (BMI) are useful actuators and diagnostic devices in applications such as intra-cardiac defibrillators, implantable pacemakers, and blood or gastric pressure measurement. BMIs are capable of accepting commands to self-adjust internal operations and of relaying the biological data of patients (BDOP) to logging stations for prognosis and evaluation. Thus, it is important to securely transmit the BDOP and maintain the integrity of commands to the BMI. How should low-power microprocessors that support the encryption of outbound BDOP and the inbound commands to the BMI for different BMI applications be devised?

The field-programmable gate array (FPGA) hardware implementations of cryptographic algorithms allow for the swapping of encryption algorithms in operation, and provide physical security and better performance than software realizations. Alternative architectural requirements for FPGA hardware implementations of RC6, Rijndael, Serpent, and Twofish encryption algorithms exist in the literature [1]. Although the throughput measurements of the implementations of different types of FPGA architectures on encryption algorithms are well known, new metrics are required for profiling the performance of symmetric block ciphers appropriate for BMIs. The authors offer useful metrics for the evaluation of contesting symmetric encryption algorithms suitable for future biomedical implant architectures. They measure the performance of 13 encryption algorithms in simulation experiments, using one and ten kilobyte workloads of blood pressure readouts by a computer program. A microarchitectural simulator capable of modeling and revealing the performance and power, and of exchanging the node capacitance of assorted functional units within the Intel XScale processor, is used to profile the symmetric ciphers.

The authors evaluate the proportional behaviors of different ciphers, in terms of the middling and climax power utilization, overall energy cost, encryption rate and efficiency, encryption program size, and the cipher security level. The Twofish cipher produced steady average and peak power consumption, while IDEA, MISTY1, LOK191, and Rijndael yielded better average power utilization. RC6, IDEA, RC5, MISTY1, and Rijndael were the top competitive energy-efficient ciphers. Unfortunately, only RC6 and RC5 survived the encryption sampling rate of BDOP, although MISTY1, Rijndael, and Blowfish stayed competitive in producing the encryption rate appropriate for BMI devices. Furthermore, the binary size of the executable algorithms of XXTEA, 3WAY, LOK191, RC6, and RC5 required memory favorable to BMI applications. All ciphers investigated, except LOK191 and DES, exhibited the property of infeasibility for a timely successful attack.

The authors convincingly advocate the need to consider the power utilization and speed in the efficient realization of load, store, move, and logic operations for BMI microprocessors. They use graphical and statistical data to highlight the factors that affect the performance of ciphers for BMI applications, but do not address how the various cipher performance metrics should be weighted, in order to objectively target the front-runner cipher. Nevertheless, the paper offers great insights into the performance evaluation of ciphers, relevant to secure data transmission with BMIs.