Research Article
Exploring the Potential of Artificial Intelligence Model to Detect Distributed Denial of Service Attacks
@INPROCEEDINGS{10.4108/eai.23-11-2023.2343336, author={Prashant Kumar and Chitra Kushwaha and Dinesh Kumar Yadav and Solomon Raju Kota}, title={Exploring the Potential of Artificial Intelligence Model to Detect Distributed Denial of Service Attacks}, proceedings={Proceedings of the 1st International Conference on Artificial Intelligence, Communication, IoT, Data Engineering and Security, IACIDS 2023, 23-25 November 2023, Lavasa, Pune, India}, publisher={EAI}, proceedings_a={IACIDS}, year={2024}, month={3}, keywords={ddos network security cyber security machine learning deep learning dos}, doi={10.4108/eai.23-11-2023.2343336} }- Prashant Kumar
Chitra Kushwaha
Dinesh Kumar Yadav
Solomon Raju Kota
Year: 2024
Exploring the Potential of Artificial Intelligence Model to Detect Distributed Denial of Service Attacks
IACIDS
EAI
DOI: 10.4108/eai.23-11-2023.2343336
Abstract
DDoS attacks, which fall under the category of cybercrime in the contemporary scene, are simple to launch yet pose enormous consequences. DDoS attacks are classified into volumetric and exploitation-based kinds, which include denial of service, LDAP, MSSQL, UDPLag, Syn, NetBIOS, UDP, and others. To detect these attacks, numerous detection methods and machine learning techniques have been deployed. Current research focuses on improving machine learning approaches, with classifiers such as Decision Trees, Support Vector Machine (SVM), and Logistic Regression displaying improved outcomes. In certain cases, algorithms were coupled to attain greater accuracy. However, dealing with massive amounts of network data presents difficulties, necessitating significant execution time and resources for sustaining hybrid models. This study investigates a deep learning model, Deep Neural Network (DNN), to effectively forecast DDoS attacks. The investigation makes use of the CICDDoS2019 benchmark dataset, which has 88 features from which a subset of 22 important features is extracted and deep learning model is applied. The proposed model's results show a significant improvement over existing techniques in this domain involving machine learning models and data mining techniques. While it is not feasible to totally eliminate the possibility of DDoS attacks, implementing the measures outlined here can help minimize these attacks to some extent. Furthermore, it enables servers to prioritize legitimate user requests rather than becoming overwhelmed by requests from illegal sources. This implementation delivers testing accuracy of 99.39%.