A self-adaptive negative selection approach for anomaly detection

To date, negative selection algorithms that possess evolutionary features, for example, the NSMutation algorithm, require the optimal value of their strategy parameters, e.g., the mutation rate and the detector lifetime indicator, to be tuned manually. The labor required for this is too time consuming and impractical when manual trial and error is used to determine the values of the strategy parameters. A reasonable alternative is to let the evolutionary algorithm determine the settings itself by using self-adaptive techniques. This work presents an evolutionary negative selection algorithm for anomaly detection (nonstationary environments) that outperforms the NSMutation on benchmark tests by using self-adaptive techniques to mutate the mutation step size of the detectors.