GAN-AIIPot: GAN-Based Cyber Deception for Probing Attacks on IoT Devices

The Internet of Things (IoT) is an emerging technology that has transformed the global network by interconnecting internet-enabled devices, people, intelligent things, and valuable data, leading to significant advancements in various domains. As IoT devices become more interwoven into our daily lives, the security of these devices is a huge concern. Many IoT devices are connected to the internet, making them open to security threats. Researchers have been exploring new methods for detecting and mitigating cyberattacks on IoT devices. One promising approach is the use of Generative Adversarial Networks (GANs) for cyber deception. Cyber deception is a cybersecurity technique used to mislead attackers and hackers from their intended targets. GANs have shown promise in the field of cybersecurity for creating realistic synthetic data to test the security of systems. In the case of probing attacks on IoT devices, GAN-based cyber deception can be used to create fake devices/information that can mimic real IoT devices and deceive attackers into thinking that they have successfully compromised a target. This paper proposes a novel GAN-based cyber deception technique called GAN-AIIPot, which is designed for probe attacks on IoT devices. GAN-AIIPot is an extended version of AIIPot that adds a GAN model on top of the Bidirectional Encoder Representations from the Transformers (BERT) model used in AIIPot. We evaluate our approach using a publicly available IoT dataset and show that GAN-AIIPot captures more sophisticated attacks and improves session length with attackers, showing the effectiveness of the deception technique compared to the existing honeypots. We believe that such a solution can enhance the security of IoT devices and protect them from malicious actors.