Temporary networks are decentralized, self-configured networks where nodes can communicate without a fixed infrastructure. They are commonly used in military, disaster recovery and IoT applications. Each node acts as a host and router, and can forward data dynamically.
When malicious nodes over-transmit fake routing requests or packets, flooding attacks occur that overflow the network, overwhelming the network. This results in resource exhaustion, increased latency and potential network failure.
The latest work on flood attack mitigation in ad hoc networks focuses on trust-based routing, machine learning classification, and adaptive intrusion detection. Technologies such as SVM, neural networks and optimization algorithms can improve attack detection, reliability and network performance. The hybrid model further improves accuracy and reduces error alerts. Despite significant progress in mitigating such attacks in this potential, current approaches are still striving to balance detection accuracy, maintain energy efficiency and adapt to rapidly changing network conditions.
In response to these challenges, a new paper recently published proposes an energy-efficient hybrid routing protocol to classify using the CNN-LSTM/GRU model to mitigate flood attacks in MANETS. A hybrid approach integrates machine learning with routing protocols to optimize energy efficiency while preventing attacks. The model classifies nodes as trusted or distrusted nodes according to the transmission behavior of its packets, thus classifying nodes that exceed a predetermined threshold as blacklisted. Training involves extracting features from benign and malicious nodes, and classification depends on learning patterns.
To improve accuracy, the model applies CNN to feature extraction, followed by LSTM or GRU for sequence learning, and optimizes decisions in real time. This protocol eliminates malicious nodes when detecting RREQ flood attacks, ensuring energy savings. MATLAB is used to create training data sets and implement Euclid-based classification. Trust estimates use link expiration time (LET) and residual energy (RE), and nodes require a minimum trust value of 0.5 to participate in routing. Finally, the ML-based AODV protocol selects nodes with the highest trust value to optimize packet delivery and minimize rerouting.
To evaluate the proposed method, the team conducted a simulation in MATLAB R2023A to evaluate the performance of a hybrid deep learning model for flood attack detection. The simulation environment accurately models the physical layer of Mona Thermo to ensure realistic evaluation conditions. Key performance metrics are analyzed, including packet delivery ratio, throughput, routing overhead, cluster head stability time, and attack detection time.
The results show that the proposed model is superior to the existing DBN, CNN and LSTM methods. It achieves a higher packet delivery rate (96.10% for 60 nodes), improves throughput (263 kbps for 100 nodes) and lower routing overhead. Furthermore, it exhibits faster attack detection times, outperforming LSTM, CNN and DBN. Classified performance indicators further confirm their advantages, with accuracy of 95%, specificity of 90% and 100% sensitivity. These findings demonstrate the effectiveness of the model in enhancing Manet security.
The proposed hybrid deep learning model shows hope for mitigating flood attacks but with limitations. Its computational complexity increases with network size, limiting real-time use in large networks and requiring a lot of memory and processing power. Furthermore, relying on MATLAB simulations may not fully reflect the real world Manet Dynamics. Regular updates and retraining are also required to accommodate evolving attack strategies.
In summary, although hybrid models (CNN-LSTM and CNN-GRU) perform better than baseline methods, challenges such as computational overhead and evolving attacks remain.
Check Paper. All credits for this study are to the researchers on the project. Also, don’t forget to follow us twitter And join us Telegram Channel and LinkedIn GrOUP. Don’t forget to join us 75K+ ml reddit.
🚨Recommended open source AI platform: “Intellagent is an open source multi-agent framework that evaluates complex dialogue AI systems” (promoted)
Mahmoud is a PhD researcher in machine learning. He still holds
Bachelor’s and Master’s Degrees in Physical Sciences
Telecom and network systems. His current field
Research involves computer vision, stock market forecasts and depth
study. He has produced several scientific articles about people
Robustness and stability of depth of identification and research
network.
✅ [Recommended] Join our Telegram Channel
