Protecting IoT Devices through Localized Detection of BGP Hijacks for Individual Things

In this paper, we leverage the limited functionality of IoT devices and the homophily of a single home network to identify control plane attacks. We illustrate the use of privacy-preserving data analysis in machine learning to evaluate the leptokurtic distributions of routes from a single device in an individual home in a specific geographic location. Previously, route hijacking has been approached as a large-scale systems problem, requiring network service providers to take action. Route information from the edge has traditionally been considered inactionable, however, small enterprises and homeowners may be targeted for such attacks for reasons ranging from nations attacking suppliers in critical systems to simple monetization of e-crime. We describe how a single small entity can leverage large-scale historical data with their individual histories to identify these attacks. We implement our proposed method in the form of a local agent that monitors the IoT devices and services for detecting BGP hijacking as well as an agent server that utilizes global history in initializing the local agents.