Intrusion Detection System is an essential part in computer security. Researchers have proposed many methods but most of them suffer from low detection rates and high false alarm rates. In this paper, we try to tackle the class imbalance problem, increase detection rates for each class and minimize false alarms in intrusion detection system. We test the performance of seven classifiers using Bagging and AdaBoost ensemble methods. We proposed a new hybrid ensemble for intrusion detection based on Error Correcting Output Code (ECOC) approach.