Adaptive IoT Defense With Deep Q-Learning Model for DNS Spoofing Prevention in Self-Organizing Networks

Self-organizing networks (SONs) can automatically manage themselves, adapting to changes without requiring centralized control. Such systems demand adaptive defense mechanisms in response to continuously evolving threats such as Domain name systems (DNS) spoofing threats. These threats lead to catastrophic impacts on the Internet of Things (IoT) systems where heterogeneous devices are connected and communicate in the form of data interceptions and malicious transferal. This paper analyzes the impact of DNS spoofing on SONs of IoTs and presents an adaptive defense mechanism against DNS spoofing with a reinforcement learning (RL) based approach. The proposed approach uses the Deep Q-Learning (DQN) model to provide defense against DNS spoofing by continuously monitoring the network condition and learning. This helps the model to lead to an optimal policy to detect the threats. This intelligent model works on the principle of continuous learning that helps to the identification of evolving and novel threats. The proposed DQN model ensures the robust security of the network. The model is trained on an IoT dataset and shows outperforming results of 97.5% accuracy on the test data.