Three-Branch Random Forest Intrusion Detection Model

Network intrusion detection has the problems of large amounts data, numerous attributes, and different levels importance for each attribute in detection. However, random forests, results have deviations due to selection attributes. Therefore, aiming at current problems, considering increasing probability essential features being selected, a network model based on three-way selected forest (IDTSRF) is proposed, which integrates three decision branches forest. Firstly, according characteristics it proposed evaluate attributes by combining boundary entropy, using rules divide attributes; secondly, keep randomness are established, certain number randomly from candidate fields conditions; finally, training sample set formed autonomous sampling method select samples sets randomly, multiple trees trained form The experimental show that high precision recall.