A Deep Learning-based System for DDoS Attack Anticipation

Author(s)

Silva, Gabriel Lucas F. M. e ; de Neira, Anderson Bergamini ; Nogueira, Michele ; Moraes, IM ; Campista, MEM ; Ghamri-Doudane, Y ; Costa, LHMK ; Rubinstein, MG

• Document type: Journal article
• Source: 2022 IEEE LATIN-AMERICAN CONFERENCE ON COMMUNICATIONS (LATINCOM); v. N/A, p. 6-pg., 2022-01-01.

Abstract

Among various threats in the Cyberspace, Distributed Denial of Service (DDoS) attacks stand out for interrupting essential services, denying access to legitimate users, and causing financial losses. The literature presents defense mechanisms against DDoS attacks. However, there are limitations in these mechanisms' response time to prevent damage effectively because of their late detection. Hence, this work presents a system based on unsupervised deep learning to identify signals of an attack in its preparation phase. Identifying early signals of an attack aims to increase the security administrator's time to stop it. The system extracts early signals from the network traffic and processes them through a deep neural network. Performance evaluation employs as input the CTU-13 dataset that contains the traffic of two different DDoS attacks. The system had anticipated the launch of the attack 46 minutes before it effectively started. (AU)

• FAPESP's process: 18/23098-0 - MENTORED: from modeling to experimentation - predicting and detecting DDoS and zero-day attacks
• Grantee: Michele Nogueira Lima
• Support Opportunities: Research Projects - Thematic Grants