Hybrid Strategy Improved Sparrow Search Algorithm in the Field of Intrusion Detection

Aiming at the problem that Sparrow Search Algorithm(SSA) may fall into local optima and have slow convergence speed, a hybrid strategy improved sparrow search algorithm(HSISSA) is proposed in this paper, and it is applied to feature selection and model optimization of intrusion detection. First, a hybrid circle-piecewise map is proposed to initialize the population and improve the uniformity of the initial population distribution; second, merging the spiral search method in the vulture search algorithm and Levy’s flight formula to update the positions of the discoverer and scouter, respectively, to expand the population search range and enhance the search capability; and finally, the simplex method and pinhole imaging method are used to optimize the position of sparrows with poor fitness and optimal fitness, to avoid stagnation in the population search and fall into local optima. The performance of the algorithm was optimized using the aforementioned methods. The algorithm was tested on 10 classical benchmark functions and combined with Wilcoxon rank-sum test analysis to verify its effectiveness, which showed improvements in convergence speed and accuracy. Finally, it was applied to the feature selection and model optimization of intrusion detection. On average, 7.6 features and 10.1 features were retained on the CIC-IDS2017 and UNSW-NB15 datasets, respectively, and 99.5% and 96.01% accuracies were achieved. The number and accuracy of the optimized features were better than those of the original algorithm. For the DenseNet and random forest models, HSISSA achieved 99.34% and 97.22% accuracy after optimization, respectively, which improved the performance of the models. Thus, the algorithm showed a better performance than the other algorithms.