Behaviour-Based Biometrics for Continuous User Authentication to Industrial Collaborative Robots

摘要

Collaborative robots (cobots) work in close proximity with human co-workers to accomplish tasks. The proximity of working arrangements and the power required of some cobots for particular tasks means that there is significant potential for cobots to cause damage to their surroundings and people nearby. Working with cobots requires appropriate training and skill. We must ensure that co-workers access appropriate levels of service and functionality from a cobot. We would wish to stop intruders engaging with cobots but also to protect against inappropriate informal working arrangements by colleagues. In this paper, we consider the potential for users' behaviours to be used as a biometric approach to continuous user authentication. More specifically, we consider how data from a cobot's internal sensors can be used to characterise a user's physical interaction with it and serve as a reference template for authentication of that user. We seek to continuously authenticate current user behaviours against these stored characteristic templates while the cobot is being manipulated (as part of a collaborative task). Our approach, based on machine learning and a recognised trust model, can provide a sensible, practical solution to authenticate users continuously as they physically interact with a cobot. Furthermore, it makes use of data that are already maintained by the cobot as part of its general operation. Our work is the first to exploit such data.