Malicious agent detection for robot swarm security (SEtS Doctoral Studentship)

About SEtS

This is a Studentship opportunity within the Doctoral Centre for Safe, Ethical and Secure Computing (SEtS) in the Department of Computer Science at the University of York. SEtS is a key initiative which supports our strategic vision to internationally lead education and research in the engineering of safe, ethical and secure computational systems. Find out more about SEtS.

About the Project

Control algorithms for robot swarms often rely on self-organisation and emergent behaviours, so the overall collective behaviour of the swarm can be influenced by the actions of a few individual robots ^[1]. This poses a problem for security, as a robot swarm could be quickly led astray by a handful of compromised agents attempting to maliciously affect its behaviour.

While network security best practices can mitigate the risk of individual robots being compromised in the first place, this research project aims to develop a more robust approach to securing a robot swarm through an active approach to malicious agent detection. Although there have been some attempts to develop systems for detecting malicious agents ^[2-5], swarm security is generally an under-studied topic that needs further attention if robot swarms are to be deployed in real-world applications.

This research project will first explore centralised approaches to building models of "normal" behaviour (e.g. using statistical methods ^[6] or machine learning), which can then be used to detect anomalous behaviour and therefore malicious agents. The project will then tackle the harder problem of malicious agent detection under the constraints of decentralised control and local sensing/communication, whereby individual robots must monitor their neighbours' behaviour and flag any suspicious activity to each other (e.g. through local consensus voting mechanisms).

It is anticipated that the research will begin in simulation, using a multi-robot simulator such as ARGoS ^[7]. The work would then be extended to physical hardware, such as a swarm of Pi-puck robots (Raspberry Pi extended e-puck robots) ^[8], in conjunction with an overhead camera and virtual sensors ^[9], with inter-robot communication implemented via Wi-Fi or XBee mesh networking.

Familiarity with one or more of the following is desired: Python/C++, Linux, robotics, embedded systems, statistical methods, machine learning.

For more details please contact Dr Alan Millard: [Email Address Removed]

Key Dates

• Application submission period: 15 November 2022 - 15 February 2023
• Interviews: 1 March - 20 March 2023
• Notification of offers: 31 March 2023
• Deadline for accepting studentship offers: 17 April 2023

Read more about the application process and follow our step-by-step guidance for applicants: How to apply

Informal enquiries

• Project enquiries: Dr Alan Millard - [Email Address Removed]
• Application enquiries: [Email Address Removed]