Scholarlink

With the increasing prevalence of network attacks, the need for robust intrusion detection systems (IDS) in Software Defined Networks (SDN) has become crucial. In this paper, a novel approach will be proposed for designing and building an IDS in SDN by leveraging a hybrid model that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) algorithms. Our system utilizes a DDOS attack SDN Dataset for training the model. The proposed hybrid CNN-LSTM model demonstrates exceptional performance in accurately detecting network intrusions. Through extensive experimentation, remarkable results have been achieved, where the accuracy of the model reached 99.9% after 500 epochs. Additionally, the loss function was reduced to a mere 0.009%, indicating the effectiveness of our approach in accurately capturing patterns associated with network attacks. To evaluate the real-time performance of our model, the Estimated Time of Arrival (ETA) for intrusion detection is computed, which averaged 269 s. This showcases the efficiency and speed of our system in identifying potential threats, thus enabling timely responses to mitigate attacks. The findings of our study highlight the potential of hybrid CNN-LSTM models in enhancing the security of SDNs by effectively detecting network intrusions. The proposed IDS offers a robust solution for network administrators to proactively defend against evolving attack vectors and safeguard the integrity and availability of SDN infrastructures.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Building and Designing an Intrusion Detection System in Software Defined Networks Against Denial-of-Service Attacks Using Deep Learning Techniques

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Thura Jabbar Khaleel,
  • Nadia Adnan Shiltagh Al-Jamali

摘要

With the increasing prevalence of network attacks, the need for robust intrusion detection systems (IDS) in Software Defined Networks (SDN) has become crucial. In this paper, a novel approach will be proposed for designing and building an IDS in SDN by leveraging a hybrid model that combines Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) algorithms. Our system utilizes a DDOS attack SDN Dataset for training the model. The proposed hybrid CNN-LSTM model demonstrates exceptional performance in accurately detecting network intrusions. Through extensive experimentation, remarkable results have been achieved, where the accuracy of the model reached 99.9% after 500 epochs. Additionally, the loss function was reduced to a mere 0.009%, indicating the effectiveness of our approach in accurately capturing patterns associated with network attacks. To evaluate the real-time performance of our model, the Estimated Time of Arrival (ETA) for intrusion detection is computed, which averaged 269 s. This showcases the efficiency and speed of our system in identifying potential threats, thus enabling timely responses to mitigate attacks. The findings of our study highlight the potential of hybrid CNN-LSTM models in enhancing the security of SDNs by effectively detecting network intrusions. The proposed IDS offers a robust solution for network administrators to proactively defend against evolving attack vectors and safeguard the integrity and availability of SDN infrastructures.