Proposal for a new architecture for detecting intrusion clusters in IoTs

In this paper, we are proposing a new architecture to detect intrusions in an IoT environment. To achieve this, we reviewed the various works to our knowledge, relating to intrusions in IoT. As a contribution, our proposed architecture has 5 components (sensors - data storage - preprocessing and associated classification algorithms). Also To provide intelligence and evaluate the performance of our architecture, we initially implemented in our architecture the DBSCAN and K-means algorithms separately. Secondly, we proceeded with a hybridization of these algorithms (DBSCAN and K-means) . In terms of results, this hybridization allowed us to use DBSCAN to identify dense clusters of arbitrary form of data as well as intrusions in this data. As for K-means, it made it possible to refine the globular clusters found by DBSCAN in order to best detect and predict the sources of intrusion. These results show that our solution makes it possible to detect intrusions in IoTs in an efficient manner compared to the 2 algorithms (DBSSAN AND K-means) applied separately in our architecture.