Abstract:

An efficient algorithm for supporting the Intrusion Detection System is required foridentifying unauthorised access that attempts to collapse a computer network's features. MachineLearning (ML) approaches like MLP and SVM Classifiers showed higher accuracy when theadditional feature selection techniques are used. Another ML approach called Deep Learning(DL) algorithm does the feature selection, automatically to overcome the extra computation offeature selection. In this paper, DL method called Stacked Autoencoder (SA) is proposed fordetecting known network anomalies using the SNMP-MIB data. SA transforms the set of inputsto a different set of reduced outputs (encoding). Previous outputs are decoded to get the desiredoutput of n dimension identical to the initial input. The proposed DL method attains a highaccuracy of 100% and saves the extra computations and resources spent on feature selection. Theproposed model was compared with 22 ML techniques and found to outperform all other allalgorithms.Keywords: deep learning; IDS; DoS; network anomalies; SNMP-MIB.