With the vulgarization of Internet, the easy access to its resources and the rapid growth in the number of computers and networks, the security of information systems has become a crucial topic of research and development especially in the field of intrusion detection. Techniques such as machine learning and data mining are widely used in anomaly-detection schemes to decide whether or not a malicious activity is taking place on a network. This paper presents a new intrusion detection system (IDS) based on information gain criterion to select relevant features from network traffic records and a new version of support vector domain description to classify the extracted features and to detect new intrusions. Experimental evaluation on NSL-KDD, a filtered version of the original KDD99 has shown that the proposed IDS can achieve good performance in terms of intrusions detection and recognition.