A novel hybrid deep learning model enhances the security and efficiency of Mobile Ad Hoc Networks (MANETs) by effectively detecting flooding attacks. This study introduces a hybrid deep learning approach combining Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) architectures to detect flooding attacks, achieving 95% accuracy, 12% increase in packet delivery ratio, and 20% reduction in routing overhead.
Mobile Ad Hoc Networks (MANETs) have gained popularity due to their decentralized and dynamic nature, making them suitable for diverse applications. Their wireless architecture, though beneficial for flexibility, introduces significant security challenges. One of the most pressing threats to MANETs is the flooding attack, where malicious nodes inundate the network with excessive packets, impairing communication and exhausting resources.
Researchers have identified the necessity of advanced detection techniques, as traditional security solutions often fall short in the face of dynamic and decentralized network structures. Previous methods predominantly relied on basic metrics such as packet transmission rates and node density, and did not leverage the powerful capabilities of deep learning approaches.
To address these shortcomings, the research team formulated their approach using hybrid deep learning models, integrating CNN with LSTM and GRU mechanisms—two architectures known for their strength in sequence prediction tasks. The CNN component is tasked with extracting relevant features from the incoming data, effectively identifying patterns indicative of flooding attacks. The LSTM and GRU then process these features to maintain temporal dependencies and refine the analysis.
Crucially, the research introduces the DECEHGS algorithm, which combines Differential Evolution and Evolutionary Population Dynamics techniques for model optimization. This algorithm enhances the model’s convergence and performance, ensuring it adapts efficiently to varying network conditions.
Validation of the proposed system was rigorously conducted using MATLAB version R2023a. Through comparative analysis with existing methods, the hybrid model demonstrated significant improvements, including achieving 95% accuracy and enhanced performance metrics such as packet delivery ratios and reductions in routing overhead. The results indicate its potential as a superior alternative for effective intrusion detection and energy conservation.
By improving attack detection efficiency, the study addresses the urgent need for adaptable and resilient solutions capable of protecting the integrity of MANET operations. These findings not only bolster the reliability of communication systems relying on MANETs but may also extend to other dynamic network frameworks facing similar security threats.
Despite the promising results, the study acknowledges limitations. The computational complexity associated with larger networks remains a concern, as the deep learning models require considerable processing capabilities, which may hinder real-time applicability. Further research is necessary to simplify the model for deployment within resource-constrained environments.
Future investigations could focus on enhancing the model's robustness against novel attack strategies and improving its performance within real-world scenarios. An adaptive learning mechanism may enable the system to self-update, maintaining its relevance against the continuously changing threat landscapes.
Overall, this research contributes significantly to the field of network security, proposing methods capable of not only detecting but also mitigating flooding attacks, thereby ensuring the sustainability and resilience of Mobile Ad Hoc Networks.