Securing Autonomous Systems

The security of an autonomous system, which consists of autonomous entities, could be compromised as a result of a physical or electronic attack. The compromise of information assets such as a system’s processing base or its communications links could potentially place an enemy ahead in terms of the OODA (Observe, Orient, Decide and Act) loop; it is this aspect of electronic warfare, which is considered in this introductory paper, ensuring that the correct information is provided to the correct entities at the correct time. In this paper particular emphasis is placed on information security and system state trust as much work has already been done to preserve the confidentiality, integrity and availability of data whilst in storage and transit. A key challenge within autonomous systems is to achieve automatic information valuation enabling individual systems to dynamically derive the classification of any information they posses. This is particularly important when decentralised information sharing is required, ensuring that only authorised entities have access to potentially sensitive information. A high-level model for representing information value (and classification) based upon context-aware computing concepts is described. If decision autonomy can be described as delegation of authority then rationally one would only prescribe higher levels of decision autonomy to a given entity if sufficient trustworthiness can be associated with it. However how is trustworthiness ascertained? A number of techniques for achieving this are described and surveyed.