Fine Tuned Deep Neural Networks for Intrusion Detection System
Published in Journal of Network Security Computer Networks, 2020
Recommended citation: D. P. Gaikwad, Amir Mukeri. (2020). "Fine Tuned Deep Neural Networks for Intrusion Detection System." Journal of Network Security Computer Networks. 6(2). http://matjournals.in/index.php/JONSCN/article/view/5143
This paper is about improving the classification capabilities of Network Intrusion Detection System (NIDS). Here we experimented with various architectures of DNNs, from smaller/shallow neural networks to more complex networks. To avoid overfitting we made use techniques such as dropout and regularization. For tuning Hyper-parameters we used techniques such as random search, grid search, Hyperband & Bayesian Optimization. Use of these techniques gradually reduces the time required to tune the DNN as well as bounding the validation error.