Feed icon

SECReT funding opportunities


This year we have a number of scholarships available. All are: More...

Published: Feb 23, 2017 8:36:00 AM

Find a SECReT supervisor
prism apply now

2011 MRes projects viewer

On the Feasibility of Using Probably Approximately Correct Search Over BitTorrent Tracking Information

Publication date:

William Mayor

This central focus point for the protocol creates a security weak spot that can be exploited. Distrupting the tracker’s service effectively halts the running of the entire BitTorrent network. In order to improve on these concerns the Distributed Hash Table (DHT) extension has been introduced and widely adopted. The DHT spreads the responsibilities of the tracker across the network and so mitigates the problems of a disrupted tracker. Whilst this improves the security of BitTorrent it has had some unintended side effects that may weaken the system’s resistance to censorship. It is possible for an attacker to focus an attack on particular elements of the DHT and in doing so censor those elements. As an alternative to a DHT, Probably Approximately Correct (PAC) search can also be used to retrieve data from distributed networks. PAC uses an unstructured distributed peer- to-peer architecture where it is assumed that it is possible to randomly sample nodes in the network. Such an architecture is expected to be more resistant to attack. This paper aims to investigate existing BitTorrent networks in order to determine how best to apply PAC search. This study also aims to reveal the extent to which a PAC solution can emulate current information retrieval performance. It is recognised that PAC search is unlikely to be able to perform as well as existing BitTorrent solutions. In light of this the study aims to examine whether there is a reasonable compromise to be made between retrieval performance and security.

A Comparison of the Spread of Extreme Protest Behaviours Through Two Activist Networks

Publication date:

Rosemary Penny

This paper examines the phenomenon of extreme protest in the environmental and animal rights movements. While both campaigns have a wide range of people protesting for them with varying degrees of militancy, the most radical protestors from the two movements adopt different methods of fighting for their cause. Using a framework called Situational Action Theory that has been successfully applied to radicalisation, this project seeks to understand how extreme protest behaviours emerge and what factors contribute to their propagation across a social network.  It does this through the construction of agent-based models describing the two activist networks, using data obtained from the activists about the types of protest activity in which they engage, their social circles, and their moral attitudes. The differences in the results emerging from the models enable conclusions to be drawn regarding which elements in the construction caused these differences, and therefore which real-world factors contribute to differences in the spread of extreme protest behaviours through social networks.

Is High Performance Liquid Chromatography Analysis a Useful Addition to Current Geo-Forensic Analytical Techniques?

Publication date:

Georgia McCulloch

Geo-forensic evidence is a potentially useful source of information for crime reconstruction due to its near ubiquitous presence, high transferability, persistence and diversity. Soil samples can be analysed in order to establish their geographical origin or to compare and exclude items of forensic interest as having a common source, however previous studies lack forensic relevance through the use of large sample amounts and comparison of samples taken from geologically different locations over large spatial scales. Furthermore, current geo-forensic analyses are based on established geological techniques and are, therefore, hindered by the philosophical differences between the disciplines of earth sciences and forensic sciences, particularly with regard to sample preparation and data interpretation. This study addresses the lack of research into the organic components of geo-forensic samples by improving upon previous research into the analysis of soil samples by high performance liquid chromatography (HPLC). A sample preparation method and set of instrument parameters are developed such that the cost of the analysis is reduced, efficiency is increased and the sample amount required for analysis is reduced four-fold. The re-developed method allows samples collected from different locations within the same site to be correctly discriminated with 100% accuracy by both visual examination of the chromatography and the use of multivariate statistics. The results of this study suggest that HPLC offers an accurate and practical method ofcomparing soil samples based on characteristics that are independent of, and therefore complementary to, traditional geo-forensic analyses.

Time-of-Flight X-Ray Compton Scatter Imaging for Cargo Security: A Preliminary Study

Publication date:

Nick Calvert

Cargo security is of increasing global concern due to the large volume of sea based freight that requires securing. Inefficient screening processes leads to delays in the delivery of goods leading to significant losses. Current screening methods rely on low energy X-Ray backscatter and high energy transmission imaging systems, restricted to providing two dimensional images. This thesis introduces the concept of Time-of-Flight (ToF) x-ray Compton scatter imaging (CSI), an imaging modality that utilises ToF information to acquire three dimensional scatter images. A mixture of simulation and experimental work was undertaken in order to assess the feasibility of such an imaging system by considering four key aspects required for a ToF CSI system. Monte Carlo simulations were performed to generate 4 MeV and 6 MeV spectra and to suggest optimal target design for a possible x-ray source. Optimal target design was found to be dependent on a number of factors relating to photon and electron yield, and photon energy. The problems of attenuation and multiple scatter were considered by performing Monte Carlo simulations of a cargo container and measuring the ToF of photons undergoing Compton scattering. ToF simulations suggested multiple scatter reached 30% for the geometry simulated and led to blurring in the ToF measured. Attenuation led to fewer counts from more distant objects, but penetration was calculated to be at least 20 cm in plastic. The timing response of a new silicon detector was experimentally compared to that of a traditional detector ,and was found to be up to 4 times slower and therefore not suitable for ToF imaging. This thesis shows that ToF CSI has potential, however future experimental work is required to verify the Monte Carlo simulations undertaken.

'Have Gun - Will Travel’: The Movement and Re-use of Firearms in England and Wales

Publication date:

Kate Gibson

Firearms criminality in England and Wales has been decreasing steadily over the last decade. Whilst Home Office statistics identify the scale of firearms criminality in England and Wales, little is known about how many of these firearms are ‘repeat offenders’, and how these weapons are offending. The aim of this research is to explore the nature of firearms used in multiple offences, and determine whether a relationship exists between offence severity and firearm behaviour post-offence.   Since 2008, the National Ballistics Intelligence Service (NABIS) has been compiling a composite database of firearms used in multiple offences, or which are still in circulation with the potential of re-use. This dataset was interrogated to explore the nature of these firearms, including the type of weapons involved in these offences, and how they are being used by offenders. For those weapons moving between police forces, network analysis techniques were employed to identify those forces with increased influence on movement networks. To explore the relationship between offence severity and firearm behaviour post-offence, correlational techniques were used to explore interactions.  The study identified that four police forces (Metropolitan Police, West Midlands Police, Merseyside Police and Greater Manchester Police) are most heavily involved in firearms criminality, accounting for 83.8% of offences, and 81.9% of recoveries. Network analysis of firearms movement identified that whilst three of these forces are predominantly exporters of weapons, Greater Manchester Police are receivers, likely due to their strong links with Merseyside, which emerged as the primary exporter.  Analysis to explore relationships between offence severity and weapon behaviour post-offence was generally inconclusive, due to limitations with the dataset. Tentative findings however identified that firearms involved in homicides may be less likely to be used in subsequent offences and more likely to be recovered.

Speed Up Effects in Security Procedure on Delhi Metro Rail : Implications for Queuing Theory and Rail Security

Publication date:

Kartikeya Tripathi

It is the thesis of this paper that in the field of railway security planning security operations on assumptions of classic queuing theory should take into account behavior of security personnel acting as servers. Due to a number of social, human, and operational factors servers do not maintain a constant rate of throughput, independent of arrival rate of passengers. This has implications both for queuing theory, and security of railways against threat of explosives. The project was made possible through working in collaboration with security authorities at Delhi Metro Rail Corporation (DMRC) in New Delhi, India. Specifications of security procedures at metro rail networks were also discussed with Veolia RATP, which is engaged in operating Mumbai’s Metro Rail (Line 1) which starts operations in 2013.

Using Semiconducting Metal Oxide Gas Sensors to Detect Explosives - A Feasibility Study

Publication date:

William Peveler

Explosives present a very real and present threat in the world today, and detection of hidden explosive devices is a key priority for security and defence practitioners. Vapour detection is one useful tool currently in development, with much focus on emulating sniffer-dogs by developing an electronic nose. Electronic noses based on semiconducting metal oxides (SMOs) are an inexpensive, portable and sensitive technology that show great promise. They do however suffer from a lack of selectivity. This project set out to explore whether or not an SMO array could differentiate between a set of explosive marker vapours. An array of ten SMO gas sensors was fabricated, based on WO3 and In2O3 inks. Production was by a commercial screen printing technique onto substrates containing gold electrodes and a platinum heater track. Nine of the sensors were tested against seven gases, including 5 explosive markers, such as nitromethane and ammonia. The testing rig was re-engineered to allow headspace sampling from both solids and liquids. Sensitivity was improved by overlaying or admixing the oxides with two zeolites, H-ZSM5 and TS-1. Each improved responses to R−NO2 and R−OH moieties respectively. The enhancement properties of gold nanoparticles were also investigated. Finally machine learning techniques were applied, testing the selectivity of the array to four of the gases used with a support vector machine (SVM). Two implementations of an SVM algorithm were run concurrently and data classification was optimised using parameter searching and normalisation. It was shown that the algorithm was capable of good classification of the data even when information on concentration was absent - a promising start for the creation of a detection device. Dstl acted as an external consultant on some aspects of the project, and initial results were presented at Fort Halstead in February 2012.

Obstructions and Requirements for Coercion Resistance

Publication date:

Pyrros Chaidos

It is self evident that a fair election must output a correct result, based on the ballots cast. However, it is unreasonable to expect a high quality (i.e. non-garbage) output from a system that receives poor input. In the context of elections, one can argue that “garbage” input are votes that do not represent the will of the voters. Votes that are coerced via threats or bribes. As internet voting takes place in an uncontrolled environment, coercion is an important security concern. In order to defend against coercion, a voting system must give voters the chance to seemingly comply with coercion requests while in reality they are able to ignore them. This work aims to improve the framework used for coercion resistance of internet voting systems, by identifying and analysing the methods that can be used to achieve it. This is accomplished in two ways. Firstly, by summarising existing theoretical limitations regarding the design of such systems and generalising a result regarding the requirements for achieving coercion resistance to fit a wider range of voting system designs. Secondly, by evaluating and comparing deployed and academic internet voting systems. In the context of this evaluation, this work includes a two-step attack as a weaker but more practical variant of the impersonation (simulation) attack, often used in defining coercion resistance. Such an attack is relevant when the authentication system used for voting is also used for other services as in Estonia and Norway. In that case, the potential cost of identity theft may be greater than the cost of non-compliance. The consequences as well as potential countermeasures to the attack are also examined.

Self-organisational behaviour patterns in crowds within the context of crime at bus stops

Publication date:

Reka Solymosi

Objectives: Transport systems in large urban areas have been identified as conducive to creating criminogenic environments. Variables such as crowding have been associated with high rates of crime incidents in certain hot spots, but the specific dynamics involved with how passengers navigate their environment has not yet been analysed, and this study aims to fill that gap.  Methods: This study takes a novel approach to tackling a problem from the social sciences by using a laboratory experiment and precise measurement tools, applied from the domain of engineering.  Passenger behaviour while boarding a bus was analysed to develop an understanding of the local-level interactions that contribute to emergent crowd self-organizational behaviours.  Results: Results show the routine activity of ‘taking the bus’, when broken down into phases, presents greatest crowding levels while passengers board the bus, followed by the time they spend waiting at the stop, and least so when they are already aboard the bus. The effect of warning messages was also considered, finding that for the most part passengers are unable to alter their spatial behaviour whilst required to complete their primary tasks, except for the time they spend ‘close’ to fellow passengers, which they reduce significantly under credible warning conditions when compared with normal conditions.   Conclusions: Using laboratory experiments to study spatial behaviour allowed for the identification of specific elements of crowding at bus stops that create crowded environments and ways in which self-organisation changes under warning conditions.

The development of a backscatter X-ray system for cargo & vehicle screening

Publication date:

Search UCL News