Federated Learning (FL) enables multiple devices or locations to train a threat detection model together without sharing original data, thereby protecting sensitive information. However, the previous models developed for threat detection have struggled with various drawbacks, scalability issues, lack of applicability to diverse scenarios, and higher computational demands. To overcome this, the research proposes a Federated Learning Enabled-Hybrid Error Distributed Generative Adversarial Network (FL-HDGAN) model for threat detection. Additionally, the model integrates a Hybrid Error analysis-based Distributed generative adversarial network (HDGAN) for an enhanced defense mechanism, which facilitates enhanced threat detection accuracy and maintains data privacy. Besides, the blacklist table within the research model improved the defense mechanism by mitigating threats. Eventually, using the KDD Cup 1999 Dataset, the FL-HDGAN model demonstrated a notable achievement of 97.65% accuracy, 97.55% sensitivity, 97.87% specificity, and denoted a minimal error rate of 0.02, compared to the existing approaches.
Mohammed et al. (Wed,) studied this question.