Multiagent DDOS attack detection model: Optimal trained hybrid classifier and entropy-based mitigation process

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Network Security

Background:

• Distributed Denial of Service (DDoS) attacks pose a significant threat to network availability and integrity.
• Existing detection methods often struggle with accuracy and efficiency, necessitating advanced solutions.

Purpose of the Study:

• To propose and evaluate a novel multi-agent system for enhanced DDoS attack detection and mitigation.
• To investigate the efficacy of a structured multi-agent approach in improving detection accuracy and response times.

Main Methods:

• A five-stage detection model: preprocessing, feature extraction, dimensionality reduction, classification (using DBN, Bi-LSTM, Deep Maxout with WUJSO optimization), and decision making.
• Utilized modified double sigmoid normalization for preprocessing and a hybrid optimization algorithm (WUJSO) for model tuning.
• Incorporated advanced feature extraction and dimensionality reduction techniques within the multi-agent framework.

Main Results:

• The proposed multi-agent system achieved a detection accuracy of 0.953 at a 90% learning rate.
• Significantly outperformed existing methods such as Bi-GRU (0.857), DEEP-MAXOUT (0.910), Bi-LSTM (0.865), RNN (0.814), NN (0.894), and DBN (0.761).

Conclusions:

• The multi-agent system demonstrates superior effectiveness in detecting and mitigating DDoS attacks.
• The structured multi-agent approach offers a promising direction for advancing robust cybersecurity measures.