We live in a highly connected world! World Wide Web, social networks, brain networks, Railway networks, Power Distribution networks etc. are all examples of connections or say networks. All these networks can be pictorially represented as a collection of connected objects. Varied it may seem but all these networks have two common entities: nodes and edges. Here, a node is an object and an edge is a line that links objects together. In the case of a social network, the nodes are human profiles and the edges can be considered as following/ follower relationships between two profiles. Similarly, in the case of railway networks- stations are nodes and the railway lines that connect the stations are edges. 

These connections in various systems help in the flow of information, power, goods etc. However, this interconnectedness comes with a risk of system failure if one of the entities/nodes of this system malfunctions or is attacked. For example: the terror attacks of September 11 targeted a single airline yet the entire airline industry came to a standstill and the hacking of a single yet influential account would result in public  outrage and the ensuing damage in the brand value of the social network. These threats are a reminder of the fact our social media platforms, air traffic, road traffic, and power distribution infrastructure are highly vulnerable to targeted attacks. Therefore, it is important that we have strategies in place to protect the networks from adversarial attacks that can cause disruption of the entire network.

“A variety of technological networks form the backbone of modern world infrastructure, and it is very essential to build safeguards to protect these networks against both failures and targeted attacks,” says Dr Karthik Raman, Associate Professor at the Bhupat & Jyoti Mehta School of Biosciences and a core member of the Robert Bosch Centre for Data Science and Artificial Intelligence (RBCDSAI), IIT Madras.

Dr Karthik Raman and his student Mr. Sai Saranga Das worked on this problem and have come up with a strategy that makes the networks more resilient to adverse attacks. The strategy basically suggests a way of judiciously re-wiring a given network to reduce the risk of network failure due to any adverse attack. The research work has been published in an esteemed research journal Physica A: Statistical Mechanics and its Applications.

The strategy is implemented through an algorithm which suggests a way to build a spare capacity in the network so that if one node of the network is attacked, the traffic of the affected node is routed through this spare capacity, as a result of which the network keeps functioning. So basically, the algorithm takes a network whose spare capacity has to be determined as an input and gives out a modified network with added spare capacity. Also, the algorithm optimizes the cost associated with the added spare capacity. 

This spare capacity can be thought of like a spare tire kept in a car; this dormant spare capacity is used if any of the four tires (nodes) in the vehicle become inoperable during a journey. In the case of a power grid, it can be thought of as the addition of spare transmission lines between the various substations in the network, such that if a major substation becomes non-functional due to a targeted attack, the quantum of power handled by the affected substation can be rerouted through these spare transmission lines. As a result, the impact of the failure of a major substation on a country’s power grid can be kept to a minimum.

The strategy has been tested on two infrastructure networks—an air traffic network and a power distribution network and it was found that it can make the network more robust to targeted attacks. The algorithm was also highly effective in increasing the robustness of the so-called canonical scale-free networks, which are representative of many real-world networks, when compared to the existing strategies to mitigate targeted attacks on these networks.

“Through this study, we have addressed the interplay between the addition of dormant spare capacity in a network and the associated capital and operational costs. Our future course of study would be to apply our algorithm in the context of biological networks to gain potentially incisive insights about them,” says Mr. Sai Saranga Das.

Contributors

Sai Saranga Das, Karthik Raman

Article

Sai Saranga Das and Karthik Raman, Effect of dormant spare capacity on the attack tolerance of complex networks. In Physica A: Statistical Mechanics and its Applications 2022; 598, 127419.

Keywords

Targeted attacks, Spare capacity, Real-world networks, Air traffic network, Power distribution network