GAN-BASED ROBUST CYBER DEFENSE SYSTEM

Abstract

Intrusion Detection Systems (IDS) are essential tools that contribute to the establishment of computer networks
without malicious activities and unauthorized access. However, conventionally constructed IDS, which are typically based on
rule-based or classical machine learning methods, have difficulties in detecting complicated and gradually evolving
cyberattacks due to their limited generalization and high false-positive rates. As a result, this research paper is devoted to the
investigation of the possibility of employing Generative Adversarial Networks (GANs) to solve such problems and thus
achieve intelligent intrusion detection. The structure of a GANs comprises two neural networks, i.e., a generator and a
discriminator, which, by competing with each other in a minimax game, thus permit the creation of feasible synthetic data and
the enhancement of the model's capability to differentiate between normal and abnormal behavior. After the generator is
trained to produce attack scenarios and the discriminator to detect intrusions, not only can the system improve its detection
accuracy, but it also gets strengthened in terms of new, unidentified threat recognition.