M-LSTM: Multiclass Long Short-Term Memory based approach for Detection of DDoS Attacks

Abstract

Distributed Denial of Service attack is a ubiquitous menace to computer networks. In this attack, several nodes attack the server by sending huge amount of traffic. Server in unable to identify the difference in requests from malicious users and benign users and hence processes all the requests. As a result of processing attack traffic, the whole network will come to halt after sometime. In this paper, an M-LSTM model has been proposed for early detection of DDoS attacks. We demonstrate the feasibility of this model by comparing results of binary, multiclass (grouped and ungrouped) classification long-short term model on CICDDoS2019 dataset. Experimental results show that Layer-2 LSTM Multiclass grouped classification yield maximum values of Precision, Recall and F-1 Score as 98.75%, 97.5% and 98% respectively.