Although some algorithms show promise in specific areas, only algorithm 6 is consistently valid, which use the products of FIN- and SYN-LRTT means.
The algorithm uses empirically derived threshold. In general, the more successful algorithms follow similar trends of utilizing the smaller packet-pair FIN- and SYN-LRTTs to make wireless host decisions.
More algorithms have higher average accuracies on the gigabit infrastructure dataset than on the megabit infrastructure dataset.
This is mostly due to losses in gigabit LRTT classification accuracies on Meg-Backbone data.
Such losses are results of megabit-infrastructure “squeezing” limiting the separation of wireless from wired ACK-pair LRTTs.
Note that megabit infrastructure effects FIN- and SYN-LRTTs as well, but not nearly as much.