Artificial Neural Network Analysis of Some Selected KDD CUP99 Dataset for Intrusion Detection

[enter Abstract Body]Due to the growing number of intrusions in local networks and the internet, it has become so universal that institution increasingly implements many structures that investigate information technology security violations. This study aimed to process, classify and predict the intrusion detection accuracy of some selected network attacks using the artificial neural network (ANN) technique. Five important attacks, namely; Buffer overflow, Denial of Service (DoS), User to Root Attack (U2R), Remote to Local Attack (R2L) and PROBE were chosen from the KDD CUPP’99 information and intrusion identification accuracy was investigated with artificial neural network (ANN) modeling technique. Findings from the classification show that out of the procedures utilized to establish the ANN model, 27262 of the 45528 buffer overflow are classified appropriately, 7903 of the 45528 DoS attacks are arranged appropriately, 1371 of the 45528 U2R are classified appropriately, 431 of the 45528 R2L are arranged appropriately and, 8304 of the 45528 PROBE are classified appropriately. Comprehensively, about 99.1% of the training proceedings are arranged properly, equivalent to 0.9% erroneous classification while the testing specimen assisted to confirm the model with 99.1% of the attacks were appropriately arranged by the ANN equation. This support that, comprehensively, the ANN equation is precise about the classification and prediction of the five attacks investigated in this study