At UNCW, I am leading research projects with a team of undergraduate researchers and master students on fileless malware detection, obfuscated malware analysis, privacy-preserving data sharing in AI, black-box fuzzing, leveraging LLMs for reverse engineering, and deep learning-guided fuzzing.
Openings: I am looking for three outstanding student researchers to work on a "cybersecurity intersection with AI" research project, focusing on one or more of the above research topics. If interested, please send me an email with your CV. Funding opportunities available!
This research aims to develop a hybrid model that combines Generative Adversarial Networks (GANs) and Transformer architectures for detecting obfuscated malware. First, the GAN will generate synthetic obfuscated malware samples by training on the ERMDS Obfuscation Dataset, mimicking various obfuscation techniques and expanding the dataset for model training. Next, a Transformer-based malware classification model is proposed, which eliminates the decoder layer and instead uses pooling methods to aggregate encoded representations for binary classification, identifying benign or malicious inputs. The main contribution is the integration of the GAN-generated adversarial data with the Transformer model, enabling high accuracy in detecting obfuscated malware through the combination of data generation and sequence modeling.