Coupled with the explosion of number of the network-oriented applications, Intrusion Detection as an increasingly popular area is attracting more and more research efforts. Although a number of algorithms have already been presented to tackle this problem, they are unable to achieve balanced detection performance for different types of intrusion and cannot respond as quickly as expected. Employing random forests algorithm (RFA)in intrusion detection, this paper devises an improved variation - IRFA and presents an IRFA based model for intrusion detection in information exchanged through network connections. The feasibility in balanced detection and the effectiveness of this approach are verified by experiments based on DARPA data sets.