Communication Network of devices resolves different major minor problems. Secure communication increases the reliability and authenticity of the network. Many of researchers proposed different models for network security. This paper has proposed a model for intrusion detection in network. Network nodes' behavior was used for the training of Random Forest (multiple decision trees). Proposed trained model identifies the session class, hence 6 decision trees were developed. In order to reduce the training feature vector modified teacher learning based optimization algorithm was proposed. In this modified TLBO more than one teacher trains and students will learn individually. Experiment was done on real network dataset NSL200. Results shows that proposed model has improved various evaluation parameters.