Intrusion Detection using Deep Learning Approach with Different Optimization
Ashwini V. Solanke · International Journal for Research in Applied Science and Engineering Technology · 2020
Intrusion detection plays an important role in security. Various deep learning approaches are used for intrusion detection, but they suffer from certain level of problems such as, high error rate and also number of iterations to be increased for processing desired output because of that accuracy of classification system gets low. The proposed system uses convolutional neural network as a deep learning approach along with different gradient optimization methods to minimize error rate in the training process to make easier. Threshold-based feature selection is used for reducing redundant or unwanted data as preprocessing step to improve performance. The comparative analysis of different optimization methods demonstrates that, the proposed system is achieved high performance accuracy to detect intrusion in traffic. Adagrad, Adadelta, RMSProp and Adam optimization algorithms are evaluated through experiment. As per experiment point of view, an Adam gives much better results in terms of precision, recall and f-measure.