Deep Learning Method For Detecting Spoofing Attacks In Internet of Things Networks

IoT, as a global technology, is a rapidly growing concept of ICT (Information and Communications Technology) systems interoperability covering many areas of life. No matter what functions are to be performed by IoT, all devices included in such a system are connected by networks. Increasing the use of these technologies in many critical fields such as medical applications cause its vulnerability in the context of many attacks in all network layers such as spoofing attack in physical layer. Recent studies have attempted to use deep learning technology to detect the spoofing attack. In this paper, we propose a novel deep learning architecture based on a hybrid CNN-BiLSTM model for the detection of spoofing attacks in IoT medical networks. Unlike traditional methods relying on handcrafted features such as Euclidean distance or Pearson correlation, our approach automatically learns rich spatial and temporal patterns from the raw signal input. Specifically, we convert physical layer signals into time-frequency representations (spectrograms), allowing the CNN model to extract local spectral features while the BiLSTM captures sequential dependencies across time. This combination enhances the ability to detect sophisticated spoofing behaviors, even in noisy or dynamic environments. The proposed model is lightweight and can be deployed on low-resource IoT medical devices. Experiments demonstrate that our method outperforms existing approaches in terms of accuracy, robustness, and execution time.