Use this URL to cite or link to this record in EThOS:
Title: Detecting cyber-physical threats against autonomous robotic systems in routine missions
Author: Bezemskij, Anatolij
ISNI:       0000 0004 7655 8430
Awarding Body: University of Greenwich
Current Institution: University of Greenwich
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Autonomous cyber physical systems are increasingly common in a wide variety of application domains, with a correspondingly wide range of functionalities and types of sensing and actuation. At the same time, the variety and frequency of cyber attacks is increasing in correspondence with the increasing popularity and functionality of these systems, from in-vehicle driver assistance to smart city infrastructure and robotics. These technologies rely on a variety of sensors, actuating nodes and control communications. Each sensor adds context by which the autonomous system can better understand its environment, but each sensor also provides opportunities for attack, as has been observed in a variety of attacks on different systems. Cyber-physical threats are increasing significantly because society is increasingly dependent on cyber-physical and Internet of things systems and devices. Cyber-physical attacks are executed by people with different motivations, intentional or not. A robotic vehicle testbed has been built and used as a testbed to develop a methodology that is capable of identifying possible threats and their causes. The design of the robotic vehicle testbed is documented with explanations in terms of its sensors, actuators and it operates. A key goal has been to develop a methodology that can automatically characterise the behaviour of the robotic testbed and be able to identify cyber-physical threats in a real-world environment. This testbed environment has met all the requirements for the experimental scenarios that we have identified. A model to observe signal characteristics, including noise level patterns on sensor data streams and incorporating this information to characterise normal or abnormal behaviour of a robotic vehicle is introduced. Following a learning phase, where the vehicle is trained in a non-attack state on the values that are considered normal, it is then subjected to a series of different cyber attacks that have physical impact (cyber-physical attacks) and physical attacks that have cyber impact (physical-cyber attacks). The problem has been approached as a binary classification problem as to whether the robot is able to self-detect if and when it is under attack. The experimental results show that the approach is promising for most attacks that the vehicle is subjected to.
Supervisor: Anthony, Richard ; Loukas, George ; Gan, Diane Sponsor: Defence Science and Technology Laboratory, Ministry of Defence
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: QA Mathematics