Decoding efficiency of the MAP and the max-log MAP algorithm as a strategy in anomaly-based intrusion detection systems

Hidden Markov Methodology, with particular care to the parameter estimation and the training phase, represents a powerful finite state machine, suitable in various recognition problems. This paper investigated the capabilities of this methodology in anomaly-based intrusion detection. The model training is performed using ML criterion, based on the gradient method. Since the attacks recognition is considered as a decoding problem, the MAP and the max log MAP algorithms combined with gradient based method were applied. The comparison between these two decoding algorithms as a strategy in anomalybased IDS is represented as well.