Secure FedIDS - Privacy preserving IDS with ensemble deep learning approach

People are increasingly reliant on the internet for communication and various devices in their daily lives. The Internet of Things (IoT) has dramatically changed many industries by allowing more automation and easy data sharing, but it also has a high security risk because it exposes many entry points for hackers. An Intrusion Detection System (IDS) is therefore necessary to signal the occurrence of threats in this type of environment. With the advancements in machine learning and deep learning frameworks, these areas have attracted considerable attention in the field of network security. The current study introduces SecureFedIDS, a new approach to network security which utilizes a hybrid ensemble of CNN and LSTM. To solve the data privacy problem, SecureFedIDS implements Federated Learning through the Flower framework. Experimental results reveal that the methods are highly effective in terms of detection rates and precision, reaching 99.4% for binary classification and 97% for multiclass classification with a minimal number of false alarms.