IoT Network Segmentation When Sensors Fail

Abstract

This draft presents a fault-tolerant network segmentation system for the Internet of Things (IoT). When devices of the IoT malfunction or fail, recovery needs to be performed to maintain system functionalities. In modern ad hoc networks like mobile ad hoc networks (MANETs), devices usually form dynamical clusters to collaboratively handle highly diverse sensing environments. To recover cluster information when parts of the IoT are not functioning, this study develops a centroid-free network segmentation algorithm that diverts dependency on centroids into empirical-space kernel matrices. The original problem of handling nonvectorial centroids is deduced to kernel matrix estimation.