Distributed Denial of Service Attack Detection for the Internet of Things Using Hybrid Deep Learning Model

As a result of the widespread adoption of the Internet of Things, there are now hundreds of millions of connected devices, increasing the likelihood that they may be vulnerable to various types of cyberattacks. In recent years, distributed denial of service (DDoS) has emerged as one of the most destructive tools utilized by attackers. Traditional machine learning approaches are typically ineffective and unable to cope with actual traffic properties when used to identify DDoS attacks. This paper introduces a novel deep learning-based intrusion detection system, specifically designed for deployment at either the Cloud or Fog level in the IoT environment. The proposed model aims to detect all types of DDoS attacks with their specific subcategory. Our hybrid model combines different types of deep learning models, including Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM), Deep Autoencoder, and Deep Neural Networks (DNNs). Our proposed model is made up of two main levels. The first one contains different parallel sub-neural networks trained with specific algorithms. The second level uses the output of the frozen first level combined with the initial data as input. The idea behind the combination of these various types of deep neural networks is to exploit their different properties to achieve very high performance. To evaluate our model, we used the CIC-DDoS2019 dataset, which satisfies all the constraints of an intrusion detection dataset. The results obtained demonstrate that our proposed model outperformed various well-known machine learning and deep learning models in terms of the true positive rate, accuracy, false alarm rate, average accuracy, and average detection rate.