A HYBRID APPROACH FOR INTRUSION DETECTION USING K-NEAREST NEIGHBOR AND ARTIFICIAL NEURAL NETWORK
Network intrusion detection is an important process in this era due to the increase of cyber violations. In this article, a hybrid approach which utilizes K-Nearest Neighbor algorithm and Artificial Neural Network to detect intrusions, is proposed. NSL-KDD dataset was used for the study. Initially, data preprocessing was carried out. Encoding was done as the first step of the pre-process which was accomplished using one hot encoding. Then, features were inserted into feature scaling which was done using Min-max normalization. Feature reduction is the final step of the pre-process which was achieved using Principal Component Analysis. Subsequently, K-Nearest Neighbor algorithm was used as binary classifier that classify data into normal and abnormal classes. Then, the abnormal class was further classified into four major attack types using Artificial Neural Network. Finally, the model was evaluated and results show that the model has high accuracy and very low overfitting and underfitting.