IoT Malware Detection with GPT Models
Rebet Keith Jones, Marwan Omar, Derek Mohammed, Calvin Nobels, Maurice Eugene Dawson · 2023
With the proliferation of IoT devices, detecting IoT malware has become a critical challenge for cybersecurity. In this study, we propose a novel approach for IoT malware detection using code gadgets and the GPT language model. We extract code gadgets from the network traffic data of IoT devices, tokenize them using the Hugging Face Transformers library, vectorize them using the GPT model's embedding layer, and feed them into the GPT model for malware detection. We evaluate our approach on two publicly available datasets, IoT -23 and Malvis, and achieve high accuracy in malware detection, with F1-scores of 0.997 and 0.986, respectively. Our approach shows promise in detecting previously unseen malware variants and can be used to enhance the security of IoT devices.