On an Intrusion-Detection Approach Via Adaptive Prediction

We introduce the notion of using adaptive (digital) prediction in an intrusion-detection application, where stimuli due to intruders are transient signals. The basic idea is to use an adaptive digital predictor for removing correlated noise. As such, this approach assists in reducing false (nuisance) alarms in the absence of intruder stimuli. It also yields an improvement in signal-to-noise ratio when intruder stimuli are present, thereby assisting in their detection. The performance of this approach is assessed in terms of a fixed-point arithmetic simulation. Some aspects of a corresponding hardware implementation that involves a commercial microprocessor are also included.