Deep Learning-Based Intrusion Detection System Using Stacked Autoencoders for Financial Fraud Detection by using ensemble techniques (CNN, LSTM and CNN)

Abstract

This look-in to investigates the mounting threats of cyber security in IoT environments and electronic financial transactions by developing an extensive intrusion detection system (IDS). IDS employs cutting-edge deep gaining knowledge of methods, in addition to an stacked autoencoder (AE) and deep neural network (DNN). The combined AE learns autonomously the basic houses from input network information in unattended, which enhances the effectiveness of future type duties. Then, beneath the supervision of DNN, a deep noble properties come from the proper categorization of the disturbance. Assessment of KDDCUP99 and NSL-KDD facts units reveals incredible overall performance, while the CNN LSTM sets an brilliant accuracy of ninety nine.9%. The following upgrades include the research of extra report methodologies including CNN and CNN LSTM models, resulting in encouragement of accuracy progress. A user-friendly the front-stop interface created the use of a flask body makes it less difficult to have interaction and authenticate customers. This examination enhances cyber security in IoT environments and virtual transactions and demonstrates the effectiveness of deep learning and document techniques in IDS. The intelligent utilization of the device is highlighted through its easy interface, which simplifies adoption in a real environment. Future endeavors are seeking to enhance report strategies and the ability to expand systemic capabilities for a much larger deployment context.