Abstract: A feature selection algorithm can improve efficiently the detection speed and result, with irrelevant and redundant data eliminated and denoised in an intrusion detection system. Taking advantage of the algorithm, a new hybrid feature selection algorithm based on the principal component analysis (PCA) in combination with decision tree algorithm (C4.5) was proposed to develop a lightweight intrusion detection system. Verifying the proposed algorithm in detail via tests with the KDD 1999 dataset, the algorithm was proved that it is available to not only ensure the high detection rate and low false alarm rate but also improve obviously the training/testing time of the intrusion detection system. Furthermore, as a result of comparative tests, the algorithm is superior to GA-SVM algorithm in training/testing time, detection rate and false alarm rate.