Evaluating Sequential Combination of Two Light-Weight Genetic Algorithm- based Solutions to Intrusion Detection

In this work we have presented a genetic algorithm approach for classifying normal connections and intrusions. We have created a serial combination of two light-weight genetic algorithm-based intrusion detection systems where each of the systems exhibits certain deficiency. In this way we have managed to mitigate the deficiencies of both of them. The model was verified on KDD99 intrusion detection dataset, generating a solution competitive with the solutions reported by the state-ofthe-art, while using small subset of features from the original set that contains forty one features. The most significant features were identified by deploying principal component analysis and multi expression programming. Furthermore, our system is adaptable since it permits retraining by using new data.