An Efficient DDoS Attack Detecting System using Levenberg-Marquardt Based Deep Artificial Neural Network Approach for IOT
Ahmed Saeed Alzahrani

Ahmed Saeed Alzahrani*, Department of Computer Science, FCIT, King Abdulaziz University, Jeddah, Saudi Arabia.

Manuscript received on December 15, 2020. | Revised Manuscript received on January 05, 2020. | Manuscript published on January 10, 2021. | PP: 59-66 | Volume-10 Issue-3, January 2021 | Retrieval Number: 100.1/ijitee.C83560110321| DOI: 10.35940/ijitee.C8356.0110321
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: The Internet of Things model envisions the widespread interconnection and collaboration of smart devices over the present and future Internet environment. Threats and attacks against IoT devices and services are on the rise due to their rapid development. Distributed-Denial-of-Service (DDoS) attacks are one of the main dangerous malwares that attack targeted organizations through infected devices. Many mechanisms are developed for IoT devices in order to detect DDoS attacks. Nonetheless, the prevailing DDoS Attack Detection (DAD) methods involve time-delay and a lower detection rate. This paper proposed an efficient approach using the Levenberg-Marquardt Neural Network (LMDANN) algorithm for detecting the DDoS attacks in order to enhance prediction accuracy. In the proposed system, a MapReduce technique is used to eliminate the redundant copies. In addition, the Entropy-based Fisher’s Discriminate Function (ENTFDF) method was developed to reduce the features from the extracted features, and the system suggests an LMDANN algorithm to classify DDoS attack data separately from the normal data. In this, 80% of the data is used for training, and 20% of the data is used for testing. The performance of the proposed LMDANN method was evaluated in contrast to other art of state algorithms (ANN, SVM, KNN, and ANFIS) in terms of some specific qualitative performance metrics (recall, sensitivity, f-measure, specificity, precision, accuracy, and training time). The results show that the proposed detection approach can efficiently detect the DDoS attack in the IoT environment, achieving 96.35% accuracy. 
Keywords: Distributed Denial of Service (DDoS), Data Deduplication, Hadoop Distributed File System (HDFS), Feature-based MinMax (F-MinMax), Entropy-based Fisher’s Discriminant Function (ENTFDF), and Levenberg-Marquardt based Deep Artificial Neural Network (LMDANN).