IoT is one of today's major growth areas. With its low computing power and massively exchanged heterogeneous data, it is vulnerable to numerous at-tacks. In this paper, we proposed a collaborative system based on hybrid intrusion detection and blockchain technology to improve IoT devices security. Thus, we simulated a federated learning using Deep Neural Network, Flower library and CICIoT2023 dataset to detect and classify IoT network traffic as malicious or benign for zero-day attack whose signatures are not yet known and for those whose signatures are already known, we use signature-based IDS. The CICIoT2023 dataset used was divided into ten (10) samples using StratifiedKFold methods and each sample was assigned to a client. The results showed the effectiveness of our approach, with an accuracy around 98.8%. At last, we developed and deployed a smart contract on Solidity to simulate the dissemination of alerts between blockchain nodes.

IoT security, federated learning, CIDS, Blockchain