Abstract One of the most important issues in securing computer networks is an Intrusion Detection System. Intrusion detection systems are searching for malicious behavior, deviation normal patterns and attacks on computer networks are discovered. This system recognizes the type of traffic allowed for unauthorized traffic. Since the today's data mining techniques to intrusion detection in computer networks are used. In this research is provided, a method for designing an intrusion detection system based on machine learning. One of the features of neural networks and machine learning systems, training is based on the training data. In this research is used for detecting the intrusion of machine learning to learn the features of the theory of Rough property that has a higher correlation coefficient is used. To train and evaluate has used the proposed approach the KDD CUP 99 dataset. This study, the accuracy of our method compares with feature-based learning algorithm, neural network self and decision tree. The simulation results show that the proposed system has high accuracy and speed of detection based on rough theory is right