Network Resilience against Cascading Failures in the Internet of Things: Theories, Algorithms and Cyber-Security Protocol Design

Project: Research

View graph of relations


The Internet of Things (IoT) is a novel paradigm in networking that is increasingly important with the advent of sensors that can be embedded into “things” as diverse as household appliances to the Internet and mission-critical systems such as the Smart Grid (next-generation power networks). The majority of things are imbued with the ability to communicate and network. Thus, the IoT is a network of networks with many unique characteristics that requires new fundamental understanding of the interactions and inter-dependencies. Where simple connectivity rules between “things” can lead to almost limitless possibilities – the complexity of IoT interconnections rapidly outstrips our ability to unravel them. The IoT networks could mash together introducing or accelerating black swan events: catastrophic cascading failures that are unexpected but obvious in hindsight.As IoT devices collect data, relay information to one another and process information collaboratively, it is important to understand the cascading prowess of IoT. An action triggered by one device, e.g., failure, could lead to a cascade of other state changes and messages being exchanged with the other devices. Understanding this embedded interaction and cascading failure effect is important to guarantee security and reliability in IoT. Detection and identification of cascading failure sources in the IoT allows timely quarantine of the failure. For example, law enforcement agencies want to identify the perpetrators who attack the Smart Grid sensors that leads to widespread power blackout. How to identify the source of the cascading failure remains a rather unexplored and challenging problem, which is complicated by the size of the network that can lead to computational intractability. There are no existing work on this problem for a general network topology.The goal of this project is to develop new mathematical theories and algorithms to identify cascading failure sources in IoT. In particular, the project aims to develop inference with multiple observations to detect failure source as quickly as possible, and to quantify the detection performance based on cascading models that well represent IoT applications. The novelty in our research is to utilize novel links between probability theory and statistical inference on graph to quantify the detection performance for as asymptotically large size, and analyze the tradeoff between reliability, security and speed of detection. The result is a new cyber-security protocol design to enhance the network resilience of IoT in a Smart Grid by identifying risks before attackers do.


Project number9042059
Grant typeGRF
Effective start/end date1/07/149/05/18

    Research areas

  • Internet of Things,Statistical Inference,Network Resilience,Smart Grid,