Journal ID : TRKU-13-08-2020-10996
[This article belongs to Volume - 62, Issue - 08]
Total View : 394

Title : Empirical Solution For An Optimized Machine Learning Framework For Anomaly-Based Network Intrusion Detection

Abstract :

Advances in tech geared targeted at moving the society to a higher plain and sophistication with ease. The further integration of Internet to ease resource dissemination is attributed to its usage ease, ubiquity in its nature, low transaction cost and trust in channel. All these will continue to advance its popularity, usage ease and adoption. The rise thus, in adversaries threatens data integrity – and the task of detecting and use of countermeasures against intrusion remains a continuous task. Our study use a convolution neural network deep learning model to detect intrusion activities. Results shows model accurately detects malicious from genuine uncompromised packets traffic with confusion matrix yielding: TP = 43, TN = 3, FN = 5, FP = 9 for model sensitivity, specificity and accuracy. Model also displays a sensitivity of 93%, specificity of 25%, accuracy of 97% with a misclassification error rate of 4% for data inclusion that were not originally used to train model.

Full article

//