At the ACM KDD Workshop on Artificial Intelligence-Enabled Cybersecurity Analytics, we presented the work Towards Enhanced IoT Security: Advanced Anomaly Detection using Transformer Models, focusing on improving threat detection in complex and data-constrained IoT environments.
The research addresses the growing challenge of securing Internet of Things (IoT) ecosystems, where traditional anomaly detection techniques often struggle to adapt to highly dynamic traffic patterns and limited labelled data. In this work, I explored the use of fine-tuned Transformer models as a flexible and scalable approach to anomaly detection, leveraging their ability to transfer knowledge from pre-trained domains to new, data-scarce scenarios.
The proposed methodology was evaluated using the CIC IoMT 2024 dataset and validated against the Aposemat IoT-23 dataset, demonstrating promising performance improvements over traditional machine learning approaches. By enabling more accurate detection of anomalous behaviour with minimal training data, this work contributes to advancing AI-driven cybersecurity solutions for next-generation IoT systems.