The Research Institute on Science of Cyber Security (RISCS) addresses two key questions: - How to measure how secure a system is - How to provide better decision support tools The institute consists of 4 projects led by Imperial, Newcastle, RHUL and UCL.  The institute is hosted by UCL. We will discuss the need for a science of cyber security and review the activities of RISCS.  The second part of the seminar will give more details of the work done at Imperial.  We have concentrated on developing an approach to use Game Theory as a basis for decision support tools to assist system administrators in protecting their systems against cyber attack.  We will review this work and report on recent results.

Targeted attacks have become an established part of the threat landscape.  Such attacks, aimed at specific victims or groups of victims, are often highly-complex, underpinned by sophisticated modules purpose-built to infiltrate an organisation and steal sensitive information, wipe data or carry out some other specific act of sabotage.  The development and implementation of these campaigns often span several years (hence the description 'Advanced Persistent Threat').  This presentation will highlight the key trends we have seen in the development of targeted attacks and offer an insight into their likely future development.

This talk will address how profit-driven cybercriminals manage to achieve cooperation, organisation and governance when operating in an extremely low trust environment. It is well known that criminals, who operate outside the law and the protection of the state, face difficulties in governing their relations due both to the requirement of secrecy and a deficit of trust. For cybercriminals the anonymity of the Internet creates even greater challenges for assessing trustworthiness and enforcing agreements. It might be expected that cybercriminals would often act alone and not in collaboration with others. Nonetheless, in recent years cybercriminals have increasingly formed links with other criminals. The purpose of this paper is to identify mechanisms that may have emerged to allow such cooperation to be established and maintained. It focusses on a discussion of: reputation; performance; appearance; and enforcement.

In this talk we will give an overview of the current state-of-the-art in automated security protocol analysis, illustrate some milestones reached, and show some of the many open challenges.

No One is Immune to Breaches: 183 Million Accounts Compromised in Q3 2014 Large Scale Attacks Against Financial Firms, Retail Companies, and Consumers’ Personal Identities and Online Accounts Are Dominant Trends.  These were just some of the findings from the third quarter 2014 Breach Level Index (BLI) released by SafeNet, Inc., a global leader in data protection solutions.

Online social media increasingly play a role in the sophisticated tactics employed by cyber criminals. Criminals often hide behind one or more online identities (so-called digital personas) to either mask their activities or gain trust of their victims. Conversely, a single persona may be shared by a group of criminals for such purposes. Unraveling these multiple digital personas is a non-trivial problem owing to the large  amounts of text communicated in online social media and the large numbers of digital personas involved. The cognitive load for cybercrime investigators is immense – existing tools lack the sophisticated capabilities required to analyse digital personas in order to provide investigators with clues to the identity of the individual or group hiding behind one or more personas. In this talk I will discuss these challenges and highlight how advances in computational analysis of natural language can help overcome them hence providing a new and powerful tool in the arsenal of cybercrime investigators.

In recent years mobile robotics has made significant strides towards applications of real value to the public domain. One prominent example is the rapid development of a variety of technologies towards enabling autonomous driving. I will outline the recent work conducted in this area by the Oxford Mobile Robotics Group (MRG) and provide an overview of the key approaches taken towards vehicle navigation and perception to enable a Nissan Leaf to autonomously re-traverse routes it has been driven along previously. I will also present the vision behind the work to provide affordable autonomy as a service on offer rather than on demand and discuss how we envision this to lead to a large-scale adoption of our technology in the short- to mid-term future. 

A key topic in today's news is Cybersecurity, and a growing field of great interdisciplinary interest is the intersection of Cybersecurity and the Law.    This talk will provide an update to US and UK laws affecting cybersecurity and related areas such as privacy.  This will include new proposed US and UK laws on cybersecurity, comparison between jurisdictions, compliance matters, and ends with a discussion of loopholes and recommendations.    

The initiative to protect against future cyber crimes requires a collaborative effort from all types of agencies spanning industry, academia, federal institutions, and military agencies. Therefore, an information exchange framework is required to facilitate breach/patch related information sharing among the participants (firms) to combat cyber attacks. In this paper, we formulate a distributed cybersecurity information sharing game, where the firms (players) can choose independently to share or not. We analyze the game from an evolutionary game point-of- view and determine the conditions under which the players’ evolutionary stability can be achieved. From the analysis, we derive a population threshold which can be exploited by the Cybersecurity Information Exchange (CYBEX) to self-enforce the firms’ participation in the CYBEX framework and reinforce them for sharing security information. We present a distributed learning heuristic to attain the evolutionary stable strategy (ESS) for various conditions. The numerical and simulation results corroborate the theoretical analysis presented to achieve ESS.

Cyber Threat Intelligence (CTI) is perceived as the next big “thing” that can help reduce, slow down or stop the increasing number of cyber-attacks.  However, CTI means different things to different people. Should that really be the case? Despite the proliferation and the cost of threat intelligence feeds, the threat level remains high in most cases i.e. mitigations against cyber-attacks are not as effective as they could be. The question is: What are the requirements for efficient cyber threat intelligence? How we can assess its success will be explored during this seminar. We will close by discussing some of the author’s work on the topic.