Leveraging Hybrid Models for DNS Spoofing Detection and Mitigation in the Financial Industry

Threats from cyber security, especially DNS spoofing attacks have presented imminent dangers to the financial transaction and communication involved in financial sectors. In this work, we explore the performance of various hybrid deep learning approaches such as LSTM, CNN (with and without residual), LSTM-CNN, GRU-CNN in detecting DNS spoofing attack types at financial industry with proposed mitigation mechanism. We use a data-centric approach to harvest information from financial transactions and communications, extract features in our pipeline with extensive pre-processing capabilities, followed by value normalization. In our experiments, the models are trained and evaluated using accuracy, precision, recall, F1-score. Comparative analyses present LSTM-CNN, the best of all possible algorithms for both early detection and identification of DNS spoofing incidents.