Find a SECReT supervisor
prism apply now

2012 MRes projects viewer

Network Externalities and Migration: An Agent-Based Model Distinguishing Documented and Undocumented Flows

Publication date:

Miranda Simon

According to theory on migration networks, when the stock of migrants at destination reaches critical mass, it attracts future migrants by virtue of the positive network externalities it generates. These network externalities can take the form of monetary or employment search assistance or it may simply mean reducing information asymmetry through communication. At a micro level, networks facilitate the interaction individuals need to inform and finance their decision to  migrate. At a macro-level, as the migrant network evolves and its density increases, it can either attract or repel future flows.

The theory can also be extended to make distinctions between legal and undocumented migration patterns. Migrating without documents involves putting oneself through a greater amount of personal danger for an unknown or possibly inexistent reward, than legal migrants. Hence, unauthorised migrants acting rationally will only leave the country when sufficient others have migrated, regardless of status, as they face higher risks and are more dependent on others for help. The theory will be tested using an agent-based simulation, allowing us to observe the emergent collective behaviour that underlies migration and the evolution of migrant networks through time. The result will be a simulation tool that can facilitate prediction of future incoming legal and unauthorised migration based on the characteristics of the migrant network existing on their side of the border.

A relevance study determining the use of GSR upon clothing and shoes as an item of evidence

Publication date:

Michaela Regan

There is currently significant debate within the field of Forensics Science concerning the weight of scientific evidence. However, to increase the evidentiary value of such samples, there needs to be empirical experimental data to provide a knowledge base for the collection, analysis and interpretation of such evidence in a forensic setting. This is extremely important for items such as trace evidence (in particular for gunshot residue (GSR)). GSR has shown to be an ambiguous evidentiary item as little is known about the manner in which it redistributes and is reincorporated on particular items after initial transfer. This study therefore aims to address this, and undertake experimental studies to provide an empirically derived knowledge base to increase the understanding of the dynamics of GSR evidence and thereby provide a means to enhance its evidentiary value.

Automated Cargo Inspection: Exploring the use of Machine Vision in X-ray Transmission Imaging

Publication date:

Thomas Rogers

X-ray transmission imaging has been widely deployed around the world to detect potential threats and other contraband in large containers, vehicles and even people. It offers a non-destructive method of inspecting goods in transportation systems and is quickly becoming mandatory. For example, the Department of Homeland Security’s Container Security Initiative aims to scan 100% of cargo containers entering the US. However, the current technique requires the manual inspection of X-ray images, which is resource intensive, time consuming and potentially prone to human error. Therefore, this proposal makes steps towards trying to automate the image inspection stage.  Dual energy X-ray imaging allows for the separation of materials (e.g. plastics from metals) in a cargo image, which may help in the automatic detection of threats. However, two problems (known as boom wobble and direction effect) may arise that can degrade the quality of material separation. The first part of this proposal will attempt to solve these problems and to hence improve material separation. The second part of this proposal is specifically aimed at the detection of counterfeit cigarettes. The detection of counterfeit cigarettes is ideal for preliminary studies of using machine vision to detect illicit goods in X-ray images. Counterfeit cigarettes are becoming increasingly prevalent across the EU. For example, one study showed that 30.9% of cigarettes in Birmingham (UK) where counterfeit and that the total number had doubled during a one year period. Moreover, the UK Border Agency has intercepted cigarettes containing asbestos, mould and human excrement. So counterfeit cigarettes can pose serious health risks to the consumer. At the same time, cigarettes should be relatively simple to detect because they are uniform and cuboidal in structure and they have periodic features in an X-ray image.  This project is funded by Rapiscan Systems and EPSRC.

Data Communication for Underwater Sensor Networks

Publication date:

Veronika Yordanova

Networked sensors have the advantage of collaboratively exploring an area of interest. This would be beneficial for underwater applications such as real time surveillance and exploration, which are impractical with the current state of technology. However, there are some fundamental limits and challenges for enabling a robust and efficient deployment of such networks. One of them is the data transfer between nodes of the network. This project aims to explore novel approaches to improving channel characteristics in underwater communications, with a view to implementing, and testing the measurable underwater performance of the system.

Trace evidence dynamics: assessing the transfer and persistence of microbial diatom evidence in forensic investigation

Publication date:

Kirstie Scott

Recognising and recovering minute quantities of evidentially significant material, can aid in forensic interpretation and provide scientific weight in criminal trial. The forensic use of diatoms has so far been focused upon diagnosing cases of drowning through identifying and attempting to exclude diatoms found in internal organs and the bone marrow with distinct localisations (e.g. Cameron, 2004; Ludes et al, 1999). Use of diatoms as trace evidence has been used to link people and objects with crimes involving freshwater and saltwater environments (e.g. Siver et al, 1993). Although diatoms naturally exist in abundance and can help to identify and compare/exclude localities, people, and objects; analysis of microbial evidence is so far underused in forensic contexts, with very little in the published literature. Diatoms are important environmental indicators due to their diversity and specificity to habitat location, and their resistance to chemical and temporal change. Diatoms primarily exist aquatically, however terrestrial and aerophilic diatoms are also dominant microscopic features of environments. The contribution of diatoms to forensic science has been researched within the realm of water; however there are significant gaps in current research as to the value of terrestrial diatoms in soils and other terrestrial surfaces, and their evidence dynamics in transferring and persisting over time.  This project aims to assess the degree of transferability of trace diatom particulates between a range of natural terrestrial habitats and their persistence on clothing over time. This study will consider the impact of variables including seasonality, moisture, and temporal decay; in order to gain further insights into the dynamics of trace microbial diatom evidence and their potential application to forensic enquiry.

Statistical change point detection of internet traffic

Publication date:

Alex Gibberd

By any measure, the UK is one of the world’s top internet users. UK dependence on cyberspace is significant and growing; more than 7% of the UK's GDP is generated through on-line activity. Networked systems are increasingly being targeted by sophisticated cyber-criminals and hostile nations. As highlighted by the Defense Select Committee, the threat that these cyber-attackers pose can "evolve at almost unimaginable speed". Since current methods are primarily based on prior knowledge of attack scenarios, these new adaptive (“zero-day”) attacks are very hard to mitigate or even detect.  This project will examine the application of advanced statistical and machine learning ideas to model and detect anomalous computer network activity, which may be relevant for detecting such attacks. Many previous efforts have only considered independent analysis of individual measures of activity. When multivariate analysis is considered it is important to establish and automatically learn the cross correlation structure present under normal operating conditions. However, this is only tractable when appropriate prior knowledge is leveraged. A common strategy in other fields is to assume that the multivariate signal forms certain patterns of clusters. To this end, recent machine learning approaches utilising sparse structure learning through the Lasso, graphical Lasso, and other extensions such as the group Lasso will be considered.  Future research may lead to the extension of this project work, incorporating the above techniques within a context-aware, multi-scale framework for anomaly detection.

Confirmation bias: A Study of biasability within Forensic anthropological visual assessments on skeletal remains

Publication date:

Sherry Nakhaeizadeh

The potential for bias in forensic science is increasingly being demonstrated, with recent studies on the issues around cognitive processes and biasability with a main focus on DNA, ballistics and fingerprints disciplines. The National Academy of Science NAS report, “A path forward” has highlighted this issue suggesting practitioners in disciplines working in the forensic sphere relying on human interpretation may be prone to error or bias. The report notes empirical research supports evidence of the effects in some disciplines, and research indicates that human error due to cognitive patterns can influence and cause forensic experts to lose their objectivity. In many disciplines such as Forensic anthropology the presence of bias, its impact, and how to mitigate its effects are still not fully assessed or appreciated, with limited research has been conducted to test the impartial judgment of the anthropologist within visual methodologies. The anthropological methods are acknowledged for being highly limited because of their subjective nature, hence for the area is in need for further research. How can we examine for bias in forensic biological anthropological profiling and thus avoid such errors that might arise from it?

The evaluation of geochemical analysis methods for forensic provenance and interpretation

Publication date:

Kelly Cheshire

The aim is to establish if it is possible to identify whether soil/sediment samples are of a mixed provenance and if so can the origin of different component elements be located with the use of analytical chemistry. In particular the homogenisation preparation step has been identified as an area to be addressed in the literature as it can have major implications on the interpretation of the results obtained as it has been demonstrated to have the potential to lead to false positive and false negative interpretations.

The detection of clandestine methamphetamine laboratories using semiconducting metal oxide gas sensors

Publication date:

David Pugh

Methamphetamine is a psychoactive amphetamine, widely sold illegally as ‘crystal meth’ in the USA, Canada and Europe.  The drug is synthesized using one of  three, relatively simple methods, meaning that marketable quantities of the compound can be produced using little equipment or knowledge of synthetic organic chemistry. This has lead to methamphetamine laboratories being formed in rented accommodation, hotels, cars, mobile homes, schools and garden sheds. Despite the significant danger posed by these laboratories to the general population, the environment, and local infrastructure, clandestine laboratories are rarely discovered through proactive detection, but are more commonly discovered accidentally or as the result of a fire or explosion. This study aims to produce an array of SMO gas sensors, known as electronic noses, to detect 8 gases commonly found in the synthesis of methamphetamine.  This array of gas sensors could be used to produce a covert device for use in hotel rooms, rented accommodation and self-storage units to detect the illegal production of methamphetamine.

Domain Adaptation of Statistical Classifiers for Security-related Bug Reports

Publication date:

Ziyu Wang

In open source software, fixing bugs depends on efforts from both public and original developers. Bug report is a text document describing details of errors or mistakes in software. It is necessary to classify the bug reports generated during this process in order to protect the security related bugs from being exploited by hackers. Compared with manual classification process, automatic classification saves time and resources.  Bug reports for different software may have different format, structure and meanings. Therefore, a classifier trained from certain database of bug reports may have low classification accuracy when applied to data from another software environment. This is regarded as a domain adaptation problem. Thus we would like to adapt a classifier to be able to test different software’s bug reports while maintaining good performance. This project aims to design a classification system with domain adaptation approach to increase the classification accuracy by adapting a trained classifier to the testing environment.

Comparative study of the different feature extraction algorithms used for fingerprint identification

Publication date:

Nabanita Basu

The research work is particularly aimed at quantitative comparison of the various feature extraction techniques that could be used individually or in combination for fingerprint identification. To test the robustness of the feature extraction algorithms, I intend to use a dataset having fingerprints of varying quality and orientation.  The different feature extraction techniques that I intend to compare include, fingerprint minutiae extraction algorithm, principal component analysis and 2D Log Gabor filters. The quality of the fingerprint image greatly influences the performance of the feature extraction algorithms.  This research work aims at combining the best image processing techniques, feature extraction procedures and pattern matching algorithms that have been used in face recognition and other forensic investigation genres, to mark out a set of feature extraction techniques that could facilitate near accurate identification of low quality, crime scene fingerprints. The research work though concentrates on working with latent fingerprints, is extendable to impressed as also plastic/ visible fingerprints. This work would sure help revolutionize the entire process of fingerprint identification. The 2004 Brandon Mayfield case highlights the need for a better, more robust, statistically more accurate and hence more reliable system for pattern recognition in the forensic domain. This research work aims at addressing this particular need.

Modelling the allocation of crowd control resources

Publication date:

Judgement in UK fingermark recovery: room for development?

Publication date:

 Helen Earwaker

The field of forensic evidence analysis is currently experiencing considerable change and controversy in the UK and abroad. The publication of the NAS report in 2009 and the UK Law Commission Report in 2011 brought heavy criticism of long standing practices of the discipline, suggesting the need for considerable reform and transparency. There is currently a move towards improving the scientific robustness of forensic science, in particular within fields of evidence analysis that claim to be able to individualise, such as fingerprint evidence. Considerable research has targeted the tasks carried out by fingerprint experts, but there is a lack of research concentrating on the tasks carried out by fingerprint development officers who visualise latent prints prior to examination by experts. This study looks to examine policies, procedures and decision making within fingerprint development in the UK. The research will involve collaboration with the Home Office Centre for Applied Science and Technology and independent UK fingerprint experts and will utilise participants from Home Office police force fingerprint development laboratories. The research aims to form the basis for recommendations which will improve judgement and decision making within fingermark development and submission leading to a reduction in loss of fingermark evidence, and aiding in the transparency of fingerprint evidence in the UK.

Assessing the potential of e-noses for illicit drug detection in future drug-trafficking interdiction strategies

Publication date:

Automating the conceptual analysis of large-scale text-based subjective data sets

Publication date:

Incorporating Nanostructures to Enhance the Performance of Semiconducting Metal

Publication date:

Gwyn Evans

Detection of explosive materials used in homemade devices has become a heightened priority in recent years, prompting a large increase in related research. Developments have been made using semiconducting metal oxides gas sensors for the detection of explosive vapours, highlighting the advantages of the technology, such as low cost, good sensitivity and rapid response times to target explosives. However, there are obstacles to overcome that have so far limited practical applications in the security field. Firstly, materials currently used in gas sensors require a high operating temperature to achieve appropriate sensitivity to the target vapour. This requires a power source that is not suitable for a discreet or portable device. Secondly, the material may respond to a range of gases of varying concentrations, lacking the selectivity required to function as a specific detector.  Recent research suggests that incorporating nanostructures, such as carbon nanotubes or graphene oxide, with traditional gas sensing materials can reduce their operating temperature, thus improving suitability for practical use. An increase in sensitivity to trace gases found in homemade explosives has also been reported, along with an improved sensor recovery time. Research has shown that the addition of various zeolites to the sensing material produces a degree of selectivity towards certain gases. The project will build upon this research, fabricating hybrid gas sensing materials using nanostructures such as carbon nanotubes, zeolites, and graphene oxide

Dual-band Frequency Reconfigurable Antennas

Publication date:

Cristina Borda

Reconfigurable antennas are becoming more widely used due to the increasing number of wireless communications and new functionalities of these systems. Moreover, multiple antennas supporting different wireless bands are not a suitable solution, because of the higher demand of compact size, efficiency, low power consumption and low cost. The aim of this project is to explore new and different reconfigurable techniques for multiple wireless applications at different frequency bands and having stable radiation patterns in all working frequencies. The novel reconfigurable antenna has to be able to operate from multiple tens of MHz to a few GHz, offering stable radiation patterns in all operating frequencies, good gain and the ability to maintain higher signal to noise ratio over the whole of the operating range frequencies. In the beginning, the aim of this MRes project is to design, build and evaluate a dual-band reconfigurable antenna, and assess if it can be adapted to a smoother frequency adjustment, to achieve one-band moving in a big range of frequencies. This project is partly funded by L3-TRL Technology

To what extent do water treatment processes affect the concentration of peroxide explosives in river water?

Publication date:

Twitter and Crime: The spatio-temporal link between social-media and criminal activity

Publication date:

Alistair Leak

Twitter is an open platform for social, business, and news purposes. Within the past 2 years it has emerged as a key factor in the detection, monitoring, and reporting of crime and political uprising as witnessed in the Arab Spring and London riots. This research seeks to investigate the relationship, if any, between the use of Twitter and the occurrence of specific crime and event types. The project will seek to identify significant links between the spatio-temporal profile of events, and the use of Twitter using space-time-scan-statistics; a 3-dimensional approach to space-time clustering. In particular, the research will seek to explain the spatial and temporal dispersion. The studies objective, being to advise, and inform policing strategy in the use of social-media, and event detection and monitoring.

Search UCL News