• abnormal behavior
• autoencoder
• classifier
• cnn
• cyber attacks
• cybersecurity
• data
• dataset
• dht
• distributed hash table
• federated learning
• ids
• internet of things
• intrusion detection system
• iot
• machine learning
• omnet
• personal
• privacy
• security
• simadl
• simulation
• smart
• smart home
• unsupervised learning

Smart devices are strongly increasingly present in people's lives, which has naturally led to the creation of smart environments. However, in order to meet the growing demand for these devices quickly, two very important issues such as the security of these devices and the confidentiality of people's sensitive data have been set aside. This work aims to ensure security within the home by detecting any abnormal behavior without neglecting personal privacy. This has been obtained by developing a distributed Intrusion Detection System (IDS) for Smart Home environments that leverages Federated Learning, an innovative and privacy preserving form of Machine Learning. The proposed solution simulates a smart home network with several smart devices that can exchange information through a Distributed Hash Table (DHT). Each smart device is equipped with an unsupervised learning algorithm to be able to recognize new types of cyber attacks. The classifier has been trained and tested with SIMADL, a publicly available dataset, generated by OpenSHS, a smart home simulator for dataset generation. Finally, the obtained IDS has been compared with a traditional approach which does not guarantee personal privacy.