The microbenchmark tests were drawn from the UnixBench 4.1.0 benchmark [21] and the lmbench 2 benchmark [18] suites. These microbenchmark tests were used to determine the performance overhead of the SELinux changes for various process, file, and socket low-level operations. These benchmarks were executed on a 333MHz Pentium II with 128M RAM. The lmbench network tests ran server programs on a 166MHz Pentium with 64MB RAM. Both the client and server machines ran the same kernel for the lmbench network benchmarks so that the results show the total cost of the SELinux overhead on both systems.