Protocol Stack for Adaptable and Secure Communications

Abstract

Modern applications using network communications demand a large set of transport service requirements. Although these requirements might be initially met at the server's end, traffic variations at intermediate nodes effect the overall connection requirements. Our thesis proposes building a protocol stack so as to be able to have flexibility in selecting a more suited protocol according to the current network conditions and deliver better bandwidth utilization and QoS. In our approach the protocol to be used is dynamically decided based on network delay; bandwidth and also security requirements. The adaptation, which involved configuring the protocol from the stack, was achieved by training based learning and adaptation using neural networks. Our results demonstrate that adapting to a better protocol as the network experiences fluctuations, result in higher utilization of the available bandwidth, less packet loss and hence provides better QoS.