Models of Concurrency and Open Computing: A one day seminar to commemorate the retirement of Mike Shields.

Professor Fred Piper, Royal Holloway, University of London

Professor mark Josephs, London South Bank University

Professior Paul Lewis, University of Southampton

Professor Yong Yan, Department of Electronics, University of Kent

Dr Hugh Deighton, LogicaCMG

Doug Amos, Synplicity Inc.

Dr Russ Clark, University of Surrey

Dr Stefan Katzenbeisser, Philips Research Eindhoven

Professor Xin Yao, University of Birmingham

Professor Yun Q Shi, New Jersey Institute of Technology

Professor Yun Q Shi, New Jersey Institute of Technology

Professor Yun Q Shi, New Jersey Institute of Technology

Associate Professor Chang-Tsun Li, University of Warwick

Professor Stephen Brewster, University of Glasgow

Dr Vesna Brujic-Okretic, City University

Professor Michael Fisher, University of Liverpool

Professor Erol Gelenbe, Imperial College London

Professor Kecheng Liu, University of Reading

Professor Chunming Rong, University of Stavanger & University of Oslo, Norway

Professor Josef Kittler, University of Surrey

Professor Ahmad-Reza Sadeghi, Horst Görtz Institute for IT Security at Ruhr-University Bochum, Germany

Dr Karoline Wiesner, School of Mathematics and Centre for Complexity Sciences, University of Bristol

Dr Peter Tino, Computer Science Department, University of Birmingham

Dr Richard Zanibbi, Rochester Institute of Technology, NY, USA

Professor Leslie Smith, University of Stirling

Prof Gavin Lowe, Computing Laboratory, University of Oxford

Risto Sarvas, Visiting Research Fellow, Digital World Research Centre, University of Surrey and Research Scientist, Helsinki Institute for Information Technology HIIT, Finland

Prof Angelo Cangelosi, Adaptive Behaviour and Cognition Research Group, School of Computing and Mathematics

As custom logic circuits (e.g. field-programmable gate arrays, or FPGAs) have become larger, the limitations of conventional design flows have become more apparent. For large designs, verification by simulation is now impractical.

Most technical people end up having to make pitches – explaining their work in order to secure the next round of funding - or to get technical work turned into real business. Sometimes you know you are going to have to pitch in a formal setting, but it can also be a spontaneous opportunity from a chance meeting.

I shall briefly review the usefulness and shortcomings of readability formulas. I will then describe CLEAR, an online aid to plain writing which I am developing: it will colour-code submitted text to show the difficulty of every word and every sentence. But simplifying sentences and cutting out jargon is not enough. So the bulk of the talk will consist of advice on structuring documents to help readers find what they need and understand what they find.

An insightful look into the exciting, cutting-edge research happening right on your doorstep. Research students from the University of Surrey present a series of short talks showcasing current activities in fields ranging from nuclear physics to the latest developments in nanotechnology and communication devices. Discover how the science and technology of today will impact tomorrow's society.

The primary benefit of digital content, the ease with which it can be duplicated and disseminated, is also the primary concern when endeavouring to protect the rights of those creating the content. Copyright owners wish to deter illicit file sharing of copyrighted material, detect it when it occurs and even trace the original perpetrator.

Alan will talk about the vision for the future of the BCS, as it undergoes a major transformation. He will outline the challenges and opportunities that the BCS faces, and the strategies that the Society is adopting to increase the engagement and relevance of the BCS within industry and within the wider community in the 21st Century.

The concept of robust and semi-fragile watermarking is described for copyright protection and authentication of digital images. A number of different transforms and algorithms used for robust and semi-fragile image watermarking are reviewed in detail. Four novel robust and semi-fragile transform based image watermarking related schemes are introduced. These include wavelet-based contourlet transform (WBCT) for both robust and semi-fragile watermarking, slant transform (SLT) for semi-fragile watermarking as well as applying generalised Benford’s Law to estimate the JPEG QF, then adjust the appropriate threshold for improving semi-fragile watermarking technique.

The need for a generic and adaptable object detection and recognition method in static images is becoming a necessity today, given the rapid development of the internet and multimedia databases in general. Comparing with human vision, the computer vision is out-performed in terms of efficiency, accuracy and depth of understanding, as the computerised recognition is achieved at contextual level. In order to achieve recognition at semantic level, computer vision systems must not only be capable of recognising objects, regardless of the changes in appearance, location, and action, but also be able to interpret abstract non-observable concepts.

As the number of electronic files containing commercially sensitive or confidential information increases every day, the importance of identifying the originality of files is becoming a hot topic for enterprises as well as computing professionals. But how can this be done in a world where ever more sophisticated image manipulation tools are widely available?

Democracy depends on elections --- the people elect those to lead them and to make decisions for them. Any election is the difficult marriage of secrecy and verifiability in that we want all the votes to be secret so that no voter feels intimidated but free to vote according to her own heart and we want the election to be verifiable so that we can all rest assured that the outcome of the election does reflect the will of the people. Elections depend on people, procedures, software and hardware --- people stand for office, vote and count the votes and if, in the heat of the moment, they get a chance many of them would cheat to get ahead. To make cheating hard we have put in place procedures that have to be followed: the ballot box is shown to be empty at the start of election day and then it is sealed; ballots are cast into it one by one; at the close of the election the box is signed; it is safely transported to a counting place and only after checking signatures and lists is it opened and finally the votes are counted under close watch from election observers.

Web evangelist Henny Swan will give an overview of trends in Web standards, with practical demonstrations showcasing new technologies and what they can do.

• Opera Software: innovative products and commitment to open Web standards
• Widgets and developing for the mobile Web
• Emerging standards: HTML5, CSS3, SVG/canvas and video
• Relevance for the Web, today and tomorrow

Linguistic Steganography aims to provide techniques for hiding information in natural language texts, through manipulating properties of the text, for example by replacing some words with their synonyms. Unlike image-based steganography, linguistic steganography is in its infancy with little existing work. 

Binary Watermarking, robust to printing and scanning, is the process of imperceptibly hiding information in binary documents, typically text documents, so that the hidden information can still be recovered following the printing and scanning of a document. It presents a challenging problem, both in finding an imperceptible way to hide data within a sparse text document, and providing an embedding strategy that can handle the myriad of distortions introduced during printing and scanning. Our goal was to develop a scheme that had sufficient capacity to embed our proposed authenticating and localising watermark. Existing schemes did not provide sufficient capacity, requiring us to develop techniques to increase the embedding capacity whilst maintaining the robustness to printing and scanning.

Information Assurance solutions are usually made up of a variety of techniques which together a level of assurance to the information being used. It is currently beyond the state of the art to automate the identification, understanding and response to developing treat vectors and the operation of Information Assurance solutions over the long term demands very heavy levels of human expertise with all of the associated costs. Part of the issue here is that whereas we as the information assurance community (developers and assessors) do have a rigorous framework for the assessment of quality in certain types of high assurance components these techniques do not exist for large parts of the infrastructure on which assured components are reliant.

Animals are capable of navigating through an environment. This requires them to recognise, remember and relate positions. Spatial representation is one of the main tasks during navigation. The brain seems to have a world centric positioning system such that we remember positions of objects in relation to a reference frame. Such spatial representation is believed to have been constructed in the hippocampus and related brain areas. Place cells, head direction cells, grid cells and border cells seem to transform vestibular information to spatial information in the brain. However, when psychological studies reveal how these areas are connected together, the process of transforming vestibular information to the kind of representation seen in place cells is still in question.

Interference between two competing stimuli has been extensively studied in many research areas including attention, information processing and cognitive control. For this study, both competition and cooperation of stimuli are explained by the developed Hopfield based Stroop model within the classical colour-word Stroop effect paradigm. Competition of stimuli occurs when the task is to name the colour for an incompatible colour-word and its colour (e.g. a word RED written in green), meanwhile the cooperation among stimuli can be observed when congruence (e.g. a word RED written in red) between both facilitates the response to the colour name. The Hopfield network is chosen for several reasons; we address the Stroop phenomenon as an association problem, the competition and cooperation of Stroop stimuli meets the pattern processing nature of the Hopfield network and the recall algorithm in Hopfield is biologically realistic. We have shown that, with a relatively simple but biologically plausible neural network of a single Hopfield network, our model is also able to predict the Stroop effect in comparison to the human performance.

Second Life is a 3D multi-user role-playing online environment. Unlike other virtual worlds, created as games with set rules and stock characters, most of what goes on in Second Life is created by its users. This makes it an ideal playground for all sort of creative people. At any given time, there are no fewer than 20,000 people active in Second Life. Over a period of 60 days as many as one and a half million registered users log in. 

Considerable effort goes into specifying secure and security protocols and the equipment in which these are embodied. In most cases the specification concentrates on positive cases with very little concentration on failure modes.

This talk will concentrate on limitations that are imposed on our ability to make assertions about the security of a system where we are unable to understand the quality of the implementation. It will do so by examining the types of failure that have led to security system failures. Finally, the talk will examine some of the extant security protocols and show that these provide very little support for identifying and guaranteeing the quality of components networked together in a distributed system

On Tuesday 23 March, the Department will hold its 7th Annual PhD Conference.
The conference celebrates the work of all of our PhD students through presentations and posters, recognising their valuable contribution to computer science research. Two keynote speakers ( Professor Rudolf Hanka - University of Cambridge and David Krause - Varian Medical Systems Inc) will present motivational talks during the event.

Tony Dyhouse will discuss some standards applicable to the fields of Information Assurance and Service Delivery, illustrating areas of commonality with regard to aim and approach. 

Our invited speaker is Dr Godfried Williams from the company Intellas UK, experts in Business Intelligence Security and Intelligent Forensics using AI techniques.

A lecture delivered by Clinton Walker, Security Consultant at Logica.  

Recent media reports covering major breaches of security claimed that they might have been prevented if staff awareness of security, procedures, appropriate data handling and security controls had been more reliable. Human error has become the biggest security concern for IT directors, end users and all parties concerned with data that’s held about them.

"Enterprises must recognise that simply trusting employees will inevitably prove detrimental to their security, their risk postures and their business interests," wrote Perry Carpenter, a research director at Gartner. Vnunet.com (10th Oct 2008).

The presentation provides a brief review of current military needs together with an assessment of how image and data processing technology can be used to meet these capability gaps.

A range of examples are presented that show how image and data processing can be used within a variety of different applications ranging from airborne, maritime, and land-based platforms. These system examples are based on specific activities and programmes undertaken by Waterfall Solutions Ltd.

There are many areas of bio-inspired computing, where inspiration is taken from a biological system and 'magically' transplanted into some engineered system. 

In this talk, Jon Timmis will explore thoughts on a slightly more principled approach to bio-inspired system development, that hopefully does not include any magic, and discuss in the context of immune-inspired systems, some of the potential and pitfalls of using biological systems as inspiration. To help ground the talk, he will explore a case study from their recent work with DSTL in the development of an immune-inspired robotic sniffer dog detection system, inspired by a signalling mechanism in T-cells that are present in the immune system

Traditional image steganalysis techniques are conducted with respect to the entire image. In this work, we aim to differentiate a stego image from its cover image based on steganalysis results of decomposed image blocks. As a natural image often consists of heterogeneous regions, its decomposition will lead to smaller image blocks, each of which is more homogeneous

On 12th and 13th July the Department of Computing is hosting the Summer meeting of NCAF, the Natural Computation Applications Forum, a platform for exchange of ideas between academia and industry.  The special theme is: Making Sense of Data - Theory and Practice.

Despite many advances in computational intelligence, it is clear that we have yet to achieve Turing’s dream of “intelligent machinery” – machines with human-level understanding. A broader reference point for intelligent behaviour which encompasses artificial agents is now being advocated (Cristianini 2010). 

In this seminar, I will prompt discussion by asking some key questions, which include “have we forgotten Turing’s dream?” and “should we abandon human intelligence as our reference point?” And of course, I will also provide my own opinion as to what I think the way forward might be.

Biologically inspired computing studies the properties and mechanisms of information processing in nature and embeds this knowledge into artificial systems. Due to its adaptability to wider range of applications, neural network has been of interest in many research areas. Furthermore, the growing evidences from the neuroscience field have led to evolutions of artificial neural network (ANN). From the simple McCulloch-Pitts models, ANN has now in its third generation with spiking neuron network (SNN) models. SNN based model provides more meaningful interpretation of biological neural system. However, information encoding is a major challenge as the trade off for its realism.

British Computer Society Debate

Adrian Seccombe (formerly Chief Information Security Officer, Eli Lilly), Ian Wells (Royal Surrey County Hospital), chaired by Roger Peel (Department of Computing, University of Surrey)

Copyright owners are faced with the task of limiting illicit file sharing of multimedia content. With this aim, Buyer-Seller Watermarking protocols are proposed to act as a suitable deterrent to file sharing by providing the copyright owner with adequate evidence of illegal distribution if and only if such illicit behaviour has occurred. A recent survey of BSW protocols concluded that only heuristic approaches to the security analysis of such protocols had been adopted in the literature and that formal analysis of the security of such schemes is a research direction worth pursuing.

Our meeting this month features the recent formation of a "Computing At School" Hub based around the BCS Guildford Branch and the Department of Computing at the University of Surrey.

There is an increasing realisation that the study of computing topics in schools is not laying the foundations for a lifelong appreciation of the subject, but instead just training students in skills related to particular current technologies. The BCS, Microsoft Research and several other industrial and educational organisations have created "Computing At School" (CAS; see http://www.computingatschool.org.uk/) as a vehicle to incubate improvements in the school-age study of computing, and to support teachers with high-quality resources to enable this change.

The BCS Guildford Branch is the first BCS Branch [to our knowledge!] to launch a local Computing At School Hub. This will be based in Guildford, but attracted teachers from south London and West Sussex as well as from Surrey and Hampshire to its Launch event last week. So far, there are about 10 Hubs country-wide, but we hope to promote our model to other Branches.

This month's meeting will introduce the Surrey CAS Hub to the BCS Guildford Branch, and outline the role that our industrial and commercial members can play to help it to achieve its goals. In particular, we will be exploring particular technologies that could be used to provide motivating learning experiences for mass teaching of the foundations of computing - such as the mobile phone.

When designing a legged robot a small change in one variable can have a significant effect on a number of the robot’s characteristics, meaning that making tradeoffs can be difficult. The algorithm presented here uses biologically inspired optimization techniques to identify the effects of changing various robot design variables and determine if there are any general rules which can be applied to the design of a legged robot. Designs produced by this simulation are compared to existing robot designs and biological systems, showing that the algorithm produces results which require less power and lower torque motors than similar existing designs, and which share a number of characteristics with biological systems

Zipf’s law, one of the empirical laws, was originally used to analyse the probability of occurrence of an event in mathematical fashion. For instance, it can be used to describe the relationships between the popularity rank of words and their frequency of use in a natural language. Similarly, it can be shown that there is a mathematical pattern between the size of the population in a country and the size of its cities.

The speaker at this seminar will be Mr Mohan Ahad, Assistant to Chief Technologist at MBDA, and Managing Director at Microlaunch Systems Ltd.  MBDA is Europe's largest supplier of Guided Weapons.  With reductions in defence budgets it has to look for innovative ways of developing technology for its military customers by leveraging capability from the civil sector.  

The talk will give an overview of the company, the funding mechanism available for low level technology and its ambitions for participating EU Framework Programmes. 

All are welcome so please make a note in your diary to attend.

Designing complex, self-organising systems is challenging. It requires to find local, decentralised rules for the agents which result in a good global performance of the overall system. In this talk, two approaches are presented at the example of a self-organising manufacturing system where local dispatching rules are used for decentralised scheduling.

The first approach supports a human designer by revealing the weaknesses of an examined manufacturing system. This is achieved by automatically searching for easy-to-analyse problem instances where the applied dispatching rule performs poorly.
The other approach is to generate the dispatching rules automatically by simulation-based Genetic Programming.

Substantial interest in digital watermarking over the last 15 years resulted in a considerable number of different watermarking systems that have been proposed.  However, despite this extensive literature on watermarking, not much progress has been made in tackling one of the most devastating attacks on these systems:
the original Stirmark attack introduced by Peticolas et al.

The attack has been described as the software equivalent of a high resolution print-scan attack which is part of a much larger class of random bending attacks (RBAs).

This talk introduces a theoretical framework based on approximate reasoning, it first extends Euclidean transformations into quantity space via the proposed fuzzy qualitative algebra, next, system behavior is represented by a set of automatically generated sampling data, then, data analysis methods are selected to extract features of the dataset according to individual application context, finally system behavior and corresponding data features are integrated at the system level. The framework is presented in terms of robotics, it has been adapted into applications with encouraging results such as hand gesture recognition, human motion analysis. A unified approach to human/prosthetic hand gesture recognition and results in vision/capture based human motion analysis will be reported in the talk. The framework is intended to provide a foundation towards a unified representation by “gluing” hybrid representations.

A discussion aimed at exploring how Watermarking might be able to help with the Identity Dilemmas.  All are welcome to attend so please enter this event in your diary.

Formal analysis is based on ensuring that a model preserves particular system properties. Defining the model and specifying the properties requires specialist expertise, the formal model and properties are typically written by hand, based on some informal definition. These definitions range from English written requirements, Unified Modelling Language (UML) models to Domain Specific Language (DSL) descriptions. Our work focuses on the investigation of whether the formal models can be automatically generated from their corresponding informal definition.

With the advancement of imaging devices and image manipulation soft-ware, production, development and manipulation of digital images can now be done by almost everyone.  For this reason, the task of tracking and protecting digital data (images, videos etc.) has become very difficult.  To provide adequate policing over the use of digital content, both Active and Passive security techniques are followed.  Digital watermarking is an active approach that involves pre-processing an image in order to protect it.  

The second in the Technologies and Applications seminar series, presented by Tony Dyhouse.

Tony Dyhouse will discuss some standards applicable to the fields of Information Assurance and Service Delivery; illustrating areas of commonality with regard to aim and approach. Different mechanisms for the protection of CIA will be discussed from a point of view of risk transference and third party provision of services, including a look at potential conflict of interest and how that can be addressed. Finally, a view on advancing technology and Cloud services.

Considerable effort goes into specifying secure and security protocols and the equipment in which these are embodied. In most cases the specification concentrates on positive cases with very little concentration on failure modes.

This talk will concentrate on limitations that are imposed on our ability to make assertions about the security of a system where we are unable to understand the quality of the implementation. It will do so by examining the types of failure that have led to security system failures. 

Finally, the talk will examine some of the extant security protocols and show that these provide very little support for identifying and guaranteeing the quality of components networked together in a distributed system

A brief introduction into why IT security has become increasingly important to businesses over recent years: what has driven the increasing use of IT in transactional business and why this has caused a focus on security.  We will also discuss the type of threats that business is aware of and what it is they believe they are responding to.  This we will use as the backdrop to describing some of the worst “real” incidents and how these might differ from the threat that business was preparing to meet.  We will then go on to talk about how software vendors view IT security and how this is driving their efforts to secure their products.  This will focus primarily on the approach that Microsoft have taken over recent years.

This presentation is given by Mr Paul King, Senior Security Advisor at CISCO.  Paul will give an overview of how organisations are at risk from corporate espionage - how organisations might be attacked and how they might reduce the risk. The talk will use Cisco's own organisation as an example. Paul will also discuss some of the research he is doing in this space.

The H.264/MVC standard offers good compression ratios for multi-view sequences by exploiting spatial, temporal and interview image dependencies.  The performance of this coding scheme is optimal in error-free channels, however in the event of transmission errors it leads to the propagation of the distorted macro-blocks, degrading the quality of experience of the user.  In this presentation we will review the state-of-the-art error concealment solutions and look into low complexity concealment methods that can be used with multi-view video coding.  Error resilience techniques that help error concealment will also be discussed.

The Law of Tendency to Executability states that all useful descriptions of processes have a tendency towards executability.  Attempts to rise above the perceived low abstraction level of executable code can produce increased expressive power, but the notations they engender have a tendency to become executable.  This has many consequences for software; its creation, evolution and deployment. It also has wider implications.  The automation that drives this tendency also raises fundamental questions about how human decision making can remain inside the execution loop.

Our next speaker in this series of seminars will be Professor Gary Yen from the Oklahoma State University.  All are welcome to attend.  

Evolutionary computation is the study of biologically motivated computational paradigms which exert novel ideas and inspiration from natural evolution and adaptation.  The applications of population-based heuristics in solving constrained and dynamic optimization problems have been receiving a growing interest from computational intelligence community.  Most practical optimization problems are with the existence of constraints and uncertainties in which the fitness function changes through time and is subject to multiple constraints.

This paper introduces a novel area of research to the Image Forensic field; identifying High Dynamic Range (HDR) digital images. We create and make available a test set of images that are a combination of HDR and standard images of similar scenes. We also propose a scheme to isolate fingerprints of the HDR-induced haloing artefact at “strong” edge positions, and present experimental results in extracting suitable features for a successful SVM-driven classification of edges from HDR and standard images. A majority vote of this output is then utilised to complete a highly accurate classification system.

In the past few years, semi-fragile watermarking has become increasingly important to verify the content of images and localize the tampered areas, while tolerating some non-malicious manipulations. Moreover, some researchers proposed self-restoration schemes to recover the tampered area in semi-fragile watermarking schemes. In this paper, we propose a novel fast self-restoration scheme resisting to JPEG compression for semi-fragile watermarking. In the watermark embedding process, we embed ten watermarks (six for authentication and four for self-restoration) into each 88 block of the original image. We then utilise four (44) sub-blocks' mean pixel values (extracted watermarks) to restore its corresponding (88) block's first four DCT coefficients for image content recovering. We compare our results with Li et al. and Chamlawi et al. DCT related schemes. The PSNR results indicate that the imperceptibility of our watermarked image is high at 37.61 dB and approximately 4 dB greater than the other two schemes. Moreover, the restored image is at 24.71 dB, approximately 2 dB higher than other two methods on average. Our restored image also achieves 24.39 dB, 22.98 dB 21.18 dB and 19.98 dB after JPEG compression QF=95, 85, 75 and 65, respectively, which are approximately 2.5 dB higher than other two self-restoration methods.

Digital video are widely used in today’s society due to the availability of a wide range of affordable digital video cameras with different specifications and functions. The manipulation of digital video is made simple with easily available processing tools, making it harder to trust them. This is where the role of digital forensics becomes important; to ensure the integrity of the evidence is restored. Digital forensics helps by providing some essential information about a video, such as to tracing the source of a digital video to the device that captured it. In this research, we propose a video camera identification technique based on the conditional probability features (CP Features). Specifically we focus on its performance for identification of video sources using cameras of different models. Using three cameras of different model, we demonstrate that the CP Features are able to correctly match the test video frames with their source. These findings provide a good indication that CP Features are suitable for digital video forensics.

The next in the series of NICE seminars.  Details to follow.

Dr Daniel Bush is our next seminar speaker in this series of NICE seminars.  All are welcome to attend.

Virtually all optical imaging systems introduce a variety of aberrations into an image. Chromatic aberration, for example, results from the failure of an optical system to perfectly focus light of different wavelengths. Lateral chromatic aberration manifests itself, to a first-order approximation, as an expansion/contraction of color channels with respect to one another. When tampering with an image, this aberration is often disturbed and fails to be consistent across the image. We describe a computational technique for automatically estimating lateral chromatic aberration and show its efficacy in detecting digital tampering.

Dr André Grüning will be our NICE seminar presenter this week.  Details to follow. We look forward to a good turnout for this interesting talk.

In this paper we have presented SULFA (Surrey University Library for Forensics Analysis) for benchmarking video forensics. This novel video library has been built for the purpose of video forensics specifically related to Camera identification and integrity verification. It contains original as well as forged video files, which will be freely available through university of surrey website. There are about 150 videos, collected from three camera sources, which are Canon sx220 (codec H.264), Nikon S3000 (codec MJPEG) and Fujifilm S2800HD (codec MJPEG). Each video is approximately 10 sec long with resolution of 240x320 and 30 frames per second. All videos have been shot after careful considering both temporal and spatial video characteristics. In order to present life like scenarios various complex and simple scene have been shot with and without using camera support (tripod).

Near field communication (NFC) is a standard-based wireless communication technology that allows data to be exchanged between devices (computers, TV, mobile, ...) that are a few centimeters apart. There will be a significant number of interesting applications (payment, ticketing, e-keys, mobile coupons, ...). One of these applications is mobile coupons (mCoupons), where users can get coupons from NFC mCoupon issuer (smart poster on the street or a NFC tagged newspaper) just by touching their (NFC-capable) mobiles. However, it would cause huge losses for companies if these coupons issued in uncontrolled way. Therefore, a secure protocol is needed to meet mCoupons requirements. Moreover, it must be formally verified, as all secure protocol must be, before building the system in the reality.

There is a proposed mCoupon protocol in the literature. Ali has formally analysed the protocol by using CasperFDR2. This is result in an attack founded. However, whether this attack is feasible in the really is another challenge especially with the different communication nature of NFC.

In this discussion, we build upon our novel research in High Dynamic Range (HDR) imaging for Image Forensics by comparing the 'halo' artifact with a similar artifact caused by JPEG compression, known as the 'ringing' artifact (Gibb's Phenomenon).  We briefly discuss the relationship between these two artifacts, how they appear in the Fourier transform space, and how reliable our scheme is at distinguishing between the two.  We also analyse and evaluate each step of our algorithm in order to optimise it for producing more accurate results.  Finally, a new framework for detecting the halo artifact is presented in reference to existing schemes that detect the ringing artifact, and our initial results will be discussed.

This is the last MSF group seminar we scheduled for the last semester, which was originally planned on 30th Jan but postponed to 6th Feb. After this seminar we will have a the new series of our group seminars for the new semester. This time Miss Hui Wang will report her study on a recent paper related to her research on semi-fragile watermarking for self-restoration.

Advanced optimisation techniques have been empirically successful in hundreds of applications; however their complexity often means they are not utilised by individuals from other disciplines whom do not necessarily have the correct background or time to fully understand their potential. Whilst these techniques perform exceptionally well on test problems designed to catch out simpler optimisation approaches, in practical situations they are not always necessary. This presentation presents a remarkably simple optimisation technique that has been successfully used to improve the specific efficiency and general characteristics of a novel transmission system. Results are compared to a canonical genetic algorithm and discussed in relation to practical considerations.

Lifespan distributions of populations of quite diverse species such as humans and yeast seem to surprisingly well follow the same empirical Gompertz-Makeham law, which basically predicts an exponential increase of mortality rate with age. This empirical law can for example be grounded in reliability theory when individuals age through the random failure of a number of redundant essential functional units.

However, ageing and subsequent death can also be caused by the accumulation of "ageing factors", for example noxious metabolic end products or genetic anomalies, such as self-replicating extra-chromosomal DNA in yeast.

We first show how Gompertz-Makeham behaviour arises when ageing factor accumulation follows a deterministic self-reinforcing process. We go then on to demonstrate that such a deterministic process is a good approximation of the underlying stochastic accumulation of ageing factors where the stochastic model can also account for old-age levelling off of mortality rate.

While NICE group seminars are supposed to share specific research activities, Prof Jianmin Jiang believe that it might be more useful to talk about funding to encourage share of experiences and secret weapons among colleagues in the department. There is an old Chinese saying that “I throw stones out in the hope to attract pieces of gold in”, which can be used to describe Prof Jianmin Jiang's motivation of this talk.

In this presentation, Prof Jianmin Jiang will talk about: (i) a brief introduction of European FP7 open funding schemes under call-9 (deadline 17th of April); (ii) How to bid for funding in FP7, his personal view; (iii) his personal plans and research activities.

Snoring may not only be unpleasant, but may also indicate some serious medical problems. Quite a few medical conditions are related to certain breathing sounds during sleep. Thus pattern analysis on such information becomes necessary. We present a system that is capable of recording overnight breathing sounds along with sensor data, and performs almost real time analysis using signal processing techniques. The analysis results can be used as a first indication concerning some serious medical conditions such as apnea, hypopnea and others. The system may also potentially be used for general sleep quality analysis and measurement. The first results from 400 patients show a high level of accuracy.

Also let me add that since this is a sleep related talk, and I wont be too technical, it is suitable for the people from the sleep clinic I know exists in the University.

Dr Nowill will present a view of Cyber and Cyber Security from the Telecommunications perspective, as well as covering the sort of work that professionals address in information and network security and related opportunities. This will be drawn from a wide range of experiences and examples with links to the National Cyber Security Strategy and how industry contributes, as well as how individuals may contribute at a personal level.

Reservoir computing and the liquid state machine model have received much attention in the literature in recent years. We investigate using a reservoir composed of a network of spiking neurons, with synaptic delays, whose synapses are allowed to evolve using a tri-phasic spike timing-dependent plasticity (STDP) rule. The results of using a tri-phasic STDP rule on the network properties are compared to those found using the more common exponential form of the rule. It is found that each rule causes the synaptic weights to evolve in significantly different fashions giving rise to different network dynamics.

The mental ability of an athlete is as crucial as their physical performance. Achievement in high performance sport requires an appropriate ‘state of mind’, which is trained alongside the physical activity. However, quantification of mental state is needed to identify, train and improve it through coaching. With the advent of a new generation of portable, compact EEGs we can measure the neurocognitive activity of an athlete’s brain. We present evidence suggesting that the ‘state of mind’ of an athlete can be measured and compared. Measurements are taken from youth, near elite and elite (GB team) archers investigating:

• quantification of archer EEG signals
• correlation of EEG data across shots
• correlation of EEG data across archers

Results demonstrate that there are measureable changes in EEG patterns during a shot with evidence suggesting that the patterns vary as a function of skill level, but not necessarily as a function of score.

This work was sponsored by the Surrey EPSRC KTA award and was done in collaboration with

• Matthew Casey & Alan Yau, Department of Computing, University of Surrey
• Keith M Barfoot, Alpha-Active Ltd
• Andrew Callaway, Centre for Event & Sport Research, Bournemouth University

Hierarchical Multi-Label Classification is a complex classification problem where an instance can be assigned to more than one class simultaneously, and these classes are hierarchically organized with superclasses and subclasses, i.e., an instance can be classified as belonging to more than one path in a hierarchical structure. We investigate the use of a neural network method and a genetic algorithm in this classification task, and compared their performances with other methods in the literature.

The formal assessment of surgical skills has grown in importance over recent years, with increasing evidence that unstructured systems of assessment have poor reproducibility, large inter-observer variation and lack of quantifiable measures. A paradigm shift has therefore begun, with the emergence of more objective and quantitative tools devised to complement current practice. In this work we aim to, through computer vision algorithms, determine if instruments can be tracked with sufficient accuracy during surgery so that the quantitative data obtained can be directly related to surgical performance, across a range of cataract types, surgeons, and equipment.

This will therefore potentially contribute to: feedback on dexterity performance during surgery; formative assessment of surgical performance; surgical training and through a combination of all of the above, improved patient safety. We present a robust algorithm based upon SURF point detection and optical flow that is capable of measuring instruments movement throughout the course of many operation procedures. In addition, the current experiments have shown that such measurements are able to separate different levels of surgeons based on their operation videos and estimate their surgical skills.

Steve will give a view on data storage from the inside of some of the key innovations that facilitated the growth of storage density and arguably underpinned the Information Age.

Reservoir Computing is a paradigm that has emerged during the last decade as viable model of how generic neural circuits can be applied to a range of classification and regression machine learning tasks.  Shown to deal elegantly with a range of real-world spatio-temporal signals such as speech and human motion, Reservoir Computing has demonstrated an advantage over more orthodox neural network techniques.

In this talk, we reveal our work on extending a Reservoir Computing model that can learn long-term, non-stationary data -- 50 years of weather data in our case.  To allow the neural circuit to adapt to seasonal and climatic changes over the time period, we apply a range of regulated Hebbian based plasticity rules to the neural synapses, as observed in real, biological neural networks from neuroscience.

The talk will look at what is Digital Forensics as compared to what is termed "computer forensics" and will focus on mobile devices and solid state "Flash" memory, the tools and the techniques used to recover extant and deleted data. Mark will look at a case example; the East Midlands Printer Bomb. Mark will conclude with a look at how current trends in technology could question conventional wisdom in the recovery of data from digital devices.

This research explores the impact of regional changes in conditions on the development of distinct groups in a population of diploid organisms.

In practice it involves computer modelling the growth of separate colonies of plants in different environments and developing appropriate measures of the changes in the population.

Designers of banking security systems are faced with a difficult challenge of developing technology within a tightly constrained budget, yet which must be capable of defeating attacks by determined, well-equipped criminals. This talk will summarise banking security technologies for protecting Chip and PIN/EMV card payments, online shopping, and online banking. The effectiveness of the security measures will be discussed, along with vulnerabilities discovered in them both by academics and by criminals. These vulnerabilities include cryptographic flaws, failures of tamper resistance, and poor implementation decisions, and have led not only to significant financial losses, but in some cases unfair allocation of liability. Proposed improvements will also be described, not only to the technical failures but also to the legal and regulatory regimes which are the underlying reason for some of these problems not being properly addressed.

Slides for downloading (PDF, 8.63 MB)

This research aims to use Evolutionary Algorithms (EA) to optimise Computational Fluid Dynamic (CFD) mesh for a turbulent jet. The Star-CD CFD package is used to construct, solve and post process the mesh and the Covariance Matrix Adaption Evolutionary Strategy (CMA-ES) algorithm is used to optimise the mesh. A Recurrent Neural Network (RNN) is also trained to predict converged CFD results, from un-converged data, aiming to reduce computation time when CFD simulations are needed for optimisation of either the CFD mesh or design of turbulent jets.

Results from a mesh optimisation loop were not positive, so attention was focused on training the RNN to predict converged CFD results and preliminary findings from this work have been encouraging.

In this talk we present the motivation, method, results and conclusions of all the work undertaken to date, as well as our future plans and ways in which the research can be further explored.

Users are often considered the weakest link in the security chain because of their poor security behaviour. One area with a vast amount of evidence related to poor behaviour is that of password management.

We have a pretty good idea of the extent to which this behaviour impacts on the individual user’s personal security. Unfortunately, we don’t know what the impact of this kind of behaviour by a number of organisational employees is, on a larger scale, nor do we know how best to intervene so as to improve the general security of an organisation as a whole. Current wisdom mandates the use of policies to curb insecure behaviours but it is clear that this approach has limited effectiveness. Unfortunately, no one really understands how the individual directives contained in the policies impact on the security of the eco-system. Sometimes directives have unexpected side-effects which are not easily anticipated.

It would be very difficult to answer this question in a real-life environment. I will describe a simulation engine which models an organisation with employee agents using a number of systems over an extended period. The simulation is tailorable, allowing tweaking of particular system-wide settings in order to implement policy dictats so as to determine their potential impact on the security of the organisation’s systems.

This tool supports security specialists developing policies within their organisations by quantifying the longitudinal impacts of particular rules.

The enforcement of authorization policies using cryptography has received considerable attention in recent years. Enforcement mechanisms vary in the amount of storage and the number of key derivation steps that are required in the worst case. These parameters correspond, respectively, to the number of edges and the diameter of the graph that is used to represent the authorization policy. In this talk we will consider a particular class of access control policies and the associated graphs. We then present a number of techniques for constructing a new graph that has a smaller diameter than the original graph but enforces the same authorization policy.

The talk is not really about access control or cryptography.  Rather, the problem of trade-offs in cryptographic access control gives rise to interesting constructions for reducing the diameter of directed acyclic graphs without adding too many edges.  It should be accessible to a general computer science audience.

Temporal data clustering can provide underpinning techniques for the discovery of intrinsic structures and can condense or summarize information contained in temporal data, demands made in various fields ranging from time series analysis to understanding sequential data. In the context of the treatment of data dependency in temporal data, existing temporal data clustering algorithms can be classified in three categories: model-based, temporal-proximity and feature-based clustering. However, unlike static data, temporal data have many distinct characteristics, including high dimensionality, complex time dependency, and large volume, all of which make the clustering of temporal data more challenging than conventional static data clustering. A large of number of recent studies have shown that unsupervised ensemble approaches improve clustering quality by combining multiple clustering solutions into a single consolidated clustering ensemble that has the best performance among given clustering solutions. Hence my research concentrates on ensemble learning techniques and its application for temporal data clustering tasks.

The costs of delivering healthcare are made much higher by the poor level of integration between healthcare IT systems. Attempts to define information standards for the exchange of healthcare data, by international organisations such as Health Level 7 (HL7), have been at best partially successful. I describe an approach to integration based on semantic data mapping – mapping diverse data formats onto a common UML model of information, and automatically generating data transforms – which has the potential to reduce the cost and complexity of integrating healthcare IT systems. Progress in applying this approach in the UK is described.

Source coding and encryption are the major operations required in the transmission of large amount of confidential information via a public network. Traditionally, these operations are performed independently. For example, the source sequence is first compressed using arithmetic coding; then the compressed sequence is encrypted using Advanced Encryption Standard (AES). In this seminar, two applications of nonlinear systems for joint source coding and encryption will be presented. For the lossless reconstruction of general binary sequences, a simple piecewise linear chaotic map is employed for simultaneous arithmetic coding and encryption. For lossy image compression, the integration of selective encryption in fractal image coding will be described.

When dealing with real data, clustering becomes a very complex problem, usually admitting many reasonable solutions. Moreover, even if completely different, such solutions can appear almost equivalent from the point of view of classical quality measures such as the distortion value. This implies that blind optimisation techniques alone are prone to discard qualitatively interesting solutions.

In this seminar, an alternative approach to clustering, including the generation of a number of good solutions through global optimisation, the analysis of such solutions through meta clustering and the final construction of a small set of solutions through consensus clustering is presented.

Ant traffic flow increases with growing density. This characteristic phenomenon is different from any other systems of traffic flow. In this talk, I would describe a computational model for density-independent traffic flow in ant colonies that transport to new nests. Ants have two types of swarm behaviour: emigration and foraging. A precedence model in computational ecology focused on foraging trails. However, ants move on a much larger scale during emigration. They gauge nest density by frequent close approaches among them and time the transport of colony. This density assessment behaviour known as quorum sensing has not been discussed in the context of traffic flow theory. Based on the behaviour, we model ant traffic flow that is organized without the influence of changes in population density of colonies. The proposed model predicts that density-independent ant traffic flow only depends on the frequency of mutual close approaches. I would show how to verify this estimation of our model in comparison with robust empirical data that ant experts obtained from field researches. I would indicate how to organize a study of computational ecology, and in which direction you may expect technical contributions using the proposed model.

Ability to learn from instructions or demonstrations is one of the fundamental properties of the brain that is necessary to acquire new knowledge or to develop novel skills and behavioural patterns. Although the concept of instruction-based learning has been studied for several decades now, the biological basis of this process remains unrevealed.

Some questions that need to be addressed are: where and how to search for the instruction-based learning in the brain? What is the neural representation of instructive signals? How do the biological neurons learn to generate desired outputs given these instructions?

In the talk I will discuss a biologically plausible model of supervised learning that addresses the above questions. I will demonstrate properties of the model in the context of such tasks as prediction, classification or internal representations. I will argue that supervised learning can contribute to reliable and precise spike-based information processing in the nervous system even in the presence of noise of different origin.

After an overview of the activities of the Honda Research Institute Europe, Prof Dr Bernhard Sendhoff will outline evolutionary optimization under the constraint of robustness. Different application domains will demonstrate the practical aspects of robust optimization. Systems engineering aims at integrating many different criteria into the optimal design of systems. He will argue that for biological as well as technical systems the high degree of spatial and temporal variability is an integral aspect of system design. Looking beyond robustness, Prof Dr Bernhard Sendhoff will introduce the biological concept of evolvability in the technical context of systems engineering. Prof Dr Bernhard Sendhoff will interpret evolvability as the capability of a technical system to respond to changes in the system's environment rapidly, efficiently and successfully.  

Finally, Prof Dr Bernhard Sendhoff will present a simplified application targeting the improvement of the evolvability of a technical system.

Internet voting is in the process of being adopted in Australia. The state of New South Wales recently used Internet voting on a large scale in the 2011 NSW State General Election. However there was a wide range of serious failures with the NSW iVote system in areas including transparency, scrutiny, engineering, risk assessment and oversight. In particular iVote did not provide verifiability, and moreover it experienced a number of critical incidents during the election. In this talk I will discuss these problems and the steps needed to avoid them in future. I will also give a brief overview of the plans in Victoria for Internet voting.

In this seminar, I'll talk about some do's and don'ts in PhD study, in particular when our students are required to complete their PhD thesis in no more than 3.5 years after undergraduate study, which is almost a mission impossible. This talk is based on "Useful Things to Know About PhD Thesis Research" by H.T. Kung and "10 easy ways to fail a PhD" by Matt Might, and of course my own experience and understanding of doing a PhD. Hopefully these discussions can help our PhD students to achieve a mission impossible.

CAPTCHAs are everywhere these days. Security and usability evaluation of CAPTCHA schemes is still an art rather than a science in the sense that it has to be done on an ad hoc basis and many steps have to be done manually. In this talk, the following questions will be focused: can we automate the security and usability evaluation process and if so to what extent? A new concept called Captchæcker (= Captcha + Checker) is proposed to automate the usability evaluation part based on machine learning, and to semi-automate the security evaluation part based on a dataflow programming framework called Reconfigurable Multimedia Coding (RMC, formerly known as Reconfigurable Video Coding = RVC). Some preliminary research results will be described and future work is explained.

Winner-take-all is a general rule commonly used in many applications such as machine learning and data mining. K-winners-take-all is a generalization of winner-take-all with multiple winners. Over the last twenty years, many K-winners-take-all neural networks and circuits have been developed with varied complexity and performance. In this talk, I will start with several mathematical problem formulations of the K-winners-take-all solutions via neurodynamic optimization, then present several K winners-take-all networks with reducing model complexity based on our neurodynamic optimization models. Finally, we will introduce the best one with the simplest model complexity and maximum computational efficiency. Analytical and Monte Carlo simulation results will be shown to demonstrate the computing characteristics and performance. The applications to parallel sorting, rank-order filtering, and information retrieval will be also discussed.

What do bruised fruit, the United States of America’s Declaration of Independence and crime scenes have in common? Our understanding of them has all been enhanced by the use of hyperspectral imaging.

Hyperspectral imaging cameras can determine if objects being viewed are hot or cold, wet or dry, their fat and sugar content and the presence of certain chemical elements. Therefore, it has a diverse range of applications in areas such as pharmaceuticals, food technology and homeland security. Whereas conventional colour cameras capture light in just three spectral windows, hyperspectral cameras have the ability to capture an entire section of the electromagnetic spectrum at every pixel. There are a number of different techniques for hyperspectral image capture including pushbroom and optically tuned filters. New capture techniques are also being developed.

In the past hyperspectral cameras were bulky and expensive and so were mostly used by the military for remote sensing and surveillance applications. Today’s hyperspectral cameras are almost as small as a standard video camera. These latest developments in camera technology are moving hyperspectral imaging from the aircraft and the military surveillance station to the laboratory and the production line.

The astonishing range of industries where these laboratories and factories are based emphasises the relevance and importance of the growth of hyperspectral imaging technology.

In agriculture hyperspectral imaging can be used to determine if soft fruit, such as apples, are bruised below the surface and likely to have a short shelf life. Similarly, HSI technology can be used in Biomedical Engineering to reveal the extent of burns and bruises below the skin of the human body.

Hyperspectral imaging is playing an increasingly important role in forensic technologies. The detection of fingerprints at crime scenes and the analysis of inks to detect forged documents can all be carried out using the technology. Hyperspectral imaging has even helped to bring new insights to old documents. The Library of Congress’ Preservation Research and Testing Division has carried out work on discarded drafts of the American Declaration of Independence to uncover crossed out words. This research has helped to give modern historians a deeper understanding into Thomas Jefferson’s thought process.

As hyperspectral imaging generates an entire section of the electromagnetic spectrum in real time for every pixel of an image, the sheer volume of data it produces can be enormous. Therefore, it requires large data storage and throughput; efficient data reduction algorithms; and intelligent and selective image capture to develop a complete system.

In general, the end user in a laboratory or factory only need a standalone turnkey system to solve a particular problem, such as whether a pharmaceutical product is counterfeit or the extent of bruising in the fruit. Such systems require new and state of the art image processing algorithms to reach correct decisions in real time.

The keynote will give an overview of Hyperspectral Imaging technology and its applications to information-processing tasks.

Electronic health records contain a wealth of information that has not been fully exploited. In the UK, clinical practices have been computerised since the 1990's whereas hospital episode data have been available since about 2005. The clinical informatics has gone significant advancement. It is possible to retrieve millions of patient records over time and across vendors and clinical practices.

In the first part of the talk, I will first present some challenges when processing and modelling health records. In the second part, I will present a case study based on the Quality Improvement Chronic Kidney Disease data set which contains nearly a million patient records. This case study shows how machine learning or pattern recognition techniques can be used to solve a data calibration problem, which otherwise, would have prevented the data from being used for epidemiology studies and worst, could lead to unnecessary referral of patients to specialists.

I will conclude the talk with a personal but possibly biased view of where research should be focused. There is plenty of room for contributions and opportunities for collaboration.

This talk will focus on how graph-structures can be compactly characterised using measurements motivated by diffusion processes and random walks. It will commence by explaining the relationship between the heat equation on a graph, the spectrum of the Laplacian matrix (the degree matrix minus the weighted adjacency matrix) and the steady-state random walk. The talk will then focus in some depth on how the heat kernel, i.e. the solution of the heat equation, can be used to characterize graph structure in a compact way. One of the important steps here is to show that the zeta function is the moment generating functions of the heat kernel trace, and that the zeta function is determined by the distribution of paths and the number os spanning trees in a graph. We will then explore a number of applications of these ideas in image analysis and computer vision.

There are informational replicators in the brain. Response characteristics, e.g. orientation selectivity in visual cortex, are known to be copied from neuron to neuron. Through STDP small neuronal circuits can undertake causal inference on other circuits to reconstruct the topology of those circuits based on their spontaneous activity. Patterns of synaptic connectivity are replicating units of evolution in the brain. The space of predictions is unlimited; brains do sparse search in this model space. We've shown that Darwinian dynamics is efficient for sparse search compared to algorithms that lack information transfer between adaptive units. Darwinian dynamics in the brain implements approximate Bayesian inference.

EEG signal classification is a challenging task in that the nature of the EEG data may vary from subject to subject, and change over time for the same subject. To improve classification performance, we propose to construct heterogeneous classifier ensembles, where not only the base classifiers are of different types, but they have different input features as well. The classification performance of the proposed method has been examined on Berlin BCI competition III data sets IVa. Our comparative results clearly show that heterogeneous ensembles outperform single models as well as ensembles having the same input features.

In this talk, Kadoo will be talking about his research done at Surrey Computing. It is about the application of discrete optimization to JPEG image compression. The main idea is to use global discrete optimization to achieve more advanced intra prediction of DC coefficients. The goal is to improve compression efficiency by significantly reducing the number of DC coefficients that need to be encoded. Therefore, this technique is called DC-less JPEG image compression.

The latest wave of technology development is being matched by equally significant changes in the way businesses, and organisations of all kinds, are doing business.  Could these changes make the UK a third world economy?

• This combination of change is totally unprecedented
• Application of technology is everything
• But who has the combination of skills and knowledge to take full advantage of these changes?

One thing is for certain, those who have the skills and see the opportunities will be the winners.

Water is not just the essential ingredient for life, but also a fundamental factor in the economy and security of any country. Coupled with increased population and climate change effect, the availability of food, water, and energy are the biggest challenges that the world faces. Over the next two decades water demand will exceed water supply by about 40% according to many scientific studies and reports. Food and energy shortages have also been described by the UK Government's Chief Scientific Advisor, Prof. Sir John Beddington, to create the "perfect storm" by 2030.

The provision of drinkable supplies through desalination could offer a sustainable solution to the drinking water problem but also presents a technical challenge too.

Seawater and brackish water are desalinated by thermal distillation and membrane methods such as reverse osmosis (RO) and electrodialysis. All these methods involve high operating and investment costs. RO is the most widely used desalination techniques, while thermal methods are mainly used in the Gulf countries. However, the high operating cost of RO is due to essential pre-treatment, scaling, bio-fouling and the high-energy consumption.

Novel desalination and renewable power generation membrane processes have been invented and developed at the Centre for Osmosis Research and Applications at the University of Surrey in collaboration with Modern Water plc. The modelling opportunities in membrane processes involve flow and mass transfer of Forward and Reverse Osmosis Processes as well as hydrodynamic and colloidal Interactions. The CORA team is also working on thermally driven membrane processes such as membrane distillation as well as direct contact heart transfer exchanger for both desalination and power generation.

The talk will present the principles of the desalination and power generation processes, results and the areas which needs modelling and optimisation.

The ability to provide more exciting, informative and entertaining end-user visual experience has created an enormous interest among the viewers towards 3D content. Whereas traditional 2D video is sufficient for describing details of captured scenes, 3D video can provide more realistic representation of the same scene with the additional value of depth. The success of today’s 3D video services requires that the end users meet a satisfactory level of perceptual quality. 3D cinema and 3DTV have grown in popularity in recent years. Filmmakers have a significant opportunity in front of them given the recent success of 3D films. In this talk we investigate whether this opportunity could be extended from cinema to home in a meaningful way.

“3D” perceived from viewing stereoscopic content depends on the viewing geometry. This implies that the stereoscopic-3D content should be captured for a specific viewing geometry in order to provide a satisfactory 3D experience. However, although it would be possible, it is clearly not viable, to produce and transmit multiple streams of the same content for different screen sizes. To solve the above problem, we analyze the performance of several disparity-based transformation techniques, which could be used for cinema-to-3DTV content conversion.

Human beings and many other species redirect their gaze towards targets of interest through rapid gaze shifts known as saccades, which are made approximately three to four times every second. While small saccades only rely on eye movements, larger ones result from coordinated movement of both eyes and head at the same time. Experimental studies have revealed that during saccades, the motor system manifests certain characteristics such as a stereotyped relationship between the relative contribution of eye and head to total gaze shift. Various optimality principles and several neural architectures have been suggested by researchers to explain these characteristics, but they do not involve incremental learning as a mechanism of optimization. Here, we suggest an open-loop neural controller with an adaptation mechanism which minimizes a proposed cost function. Simulations show that the characteristics of gaze shifts generated by this model match the experimental data in many aspects, in both head-restrained and head-free conditions. Therefore, our model can be regarded as a first step towards bringing together an optimality principle, a neural architecture, and an incremental learning mechanism into a unified control theory of saccadic gaze shifts.

Stegobot is a proof of concept new-generation botnet where bots communicate over unobservable communication channels. It is designed to spread via social-malware attacks and steal information from its victims. Unlike conventional botnets, Stegobot traffic does not introduce new communication endpoints between bots. Command and control information as well as stolen sensitive information are relayed using steganographic techniques piggybacking over the image sharing behavior of users in a social network. Hence stolen information travels along the edges of the victims' social network. The current implementation is based on a simple routing algorithm called restricted flooding. The tuning of the steganographic channels is a key security parameter. It works surprisingly well in real world experimental deployments; even when tuned very conservatively (against detection) it is capable of channeling sensitive payloads of close to 100MB to the botmaster. See press coverage in the New Scientist, MSNBC, Times of India, and a few others.

I will present a neurocomputational model that we developed to simulate and explain, at cortical level, word learning and language processes as they are believed to occur in motor and sensory primary, secondary and higher association areas of the (inferior) frontal and (superior) temporal lobes of the human brain. Mechanisms and connectivity of the model aim to reflect, as much as possible, functional and structural features of the corresponding cortices, including well-documented (Hebbian) associative learning mechanisms of synaptic plasticity. The model was able to explain and reconcile seemingly incongruous results on neurophysiological patterns of brain responses to well-learned, familiar sensory input (words) and new, unfamiliar linguistic material (pseudowords), and made novel predictions about the complex interactions between language and attention processes in the human brain. To test the validity of these predictions we carried out a new MEG study in which we presented subjects with familiar words and matched unfamiliar pseudowords during attention demanding tasks and under distraction. The experimental results indicated strong modulatory effects of attention on the brain responses to pseudowords, but not on those to words, fully confirming the model's predictions.

In the second part I will illustrate how the same six-area network architecture, implementing the same functional features, can be used to model and explain also cortical mechanisms underlying working memory processes, in the language as well as in the visual domain. In particular, I will present new simulation results that provide a mechanistic answer to the question of why "memory cells" (neurons exhibiting persistent activity in working memory tasks that require stimulus information to be kept in mind in view of future action) are found more frequently in prefrontal cortex and higher sensory areas than in primary cortices, i.e. far away from the  sensorimotor activations that bring about their formation (a phenomenon that we refer to as "disembodiment" of memory). The results point to the intrinsic connectivity of the sensorimotor cortical structures within which the correlation learning mechanisms operate as to the main factor determining the observed topography of memory cells.

High quality of recovered image need much more information input as watermark. However, if we want guarantee a certain level of invisibility of watermark and its robustness to some normal image processing, the capacity of watermark is limited. It’s hard to balance each other to get a good performance. This paper reports a systematic approach for benchmarking semi-fragile watermarking algorithms for authentication and restoration. The benchmarking approach is based on a generic dataflow framework of all semi-fragile authentication-restoration watermarking systems and it defines a set of performance metrics that reflect different aspects of the overall performance. Both non-malicious manipulations and attacks are modelled by a channel simulator between the sender and the receiver. The benchmarking process can be done at two levels: system level and component level. The system level benchmarking is the one normally conducted in the literature where a number of algorithms are compared without considering their internal structures. The component level benchmarking is done by reconfiguring only one component of an existing algorithm in order to observe the influence of the component to the final performance of the whole system. Following the benchmarking approach, a software system was developed to benchmark three selected semi-fragile authentication-restoration watermarking algorithms working in DCT domain and variants of the three algorithms. The benchmarking results lead to some insights about how to further improve performance of existing algorithms.

A new method called vectorized structure approach is proposed for truss topology optimization, which effectively increases or reduces the structure elements by iteratively adding or removing nodes and bars. The approach was developed on structural vectorization using nodes’ coordinates. The mapping relationship between structural vectors and realistic structures is established by the geometry composition analysis rules. Therefore, the approach can generate the optimal solution contains the given set of required nodes. Two kinds of topology optimization methods are considered. The first method sorts vector elements in every mapping process and the second method just consider sorting elements itself to be part of optimizing process. The numerical experiments have shown that both methods are effective.

Processing history recovery is a branch of digital forensic that deals with non-malicious processing such as image compression. Different types of compression leaves different artefacts which make it harder to have a universal audit trail process. This paper investigates the processing history of a JPEG2000 compressed images. We present a novel rule based compression strength estimation technique for JPEG2000 images. The technique uses a no-reference (NR) perceptual blur metric and double compressions calibration to produce a heuristic rule based algorithm, which classifies an image into three compression categories. These categories are derived subjectively and represent high, medium and low compression for a compression range of 0.1 to 0.9 bits per pixel (bpp). The technique works based on the assumption that blur artefact contained in a JPEG2000 compressed image can be linked to blur artefact contained in the calibrated double compressed image. Based on this observation, three ratios can be calculated by comparing the levels of blur of the target image and its calibrated editions, which can then be used to form three rules each corresponding to a category of JPEG2000 compression strength. For our experiments we have used 100 images to identify rules and 100 for blind testing of our scheme. Comprehensive testing showed that the compression rate of JPEG2000 images can be estimated into high, medium and low compression category with 90% accuracy.

The concept of brain connectivity is introduced to explore and understand the organized structure of cortical regions, during the execution of a specific task. Popular methods for measuring brain connectivity include synchronization, coherency and correlation among EEG sites and exploiting the statistical properties of the EEG signals to define the direction of the information flow between the cortical regions. Our approach on the problem of brain connectivity uses the theory of time-space adaptive filters. We use a network of adaptive filters that is self-organised in space and evolves through time in order to model a certain pattern of propagation of EEG signals. Exploiting the properties of these filters can lead us to a robust and flexible method, which can be used in a variety of applications.

There is an ever increasing need of identity authentication in our daily life, from making purchase online and unlocking personal devices to accessing secure premises. Although physical lock, PIN and gesture (which is common for mobile devices) offer convenient solutions, only biometrics based authentication can provide the ultimate means of validating the identity credential. The use of biometric authentication entails many questions, especially when it is used over a long period of time and in different locations: Will the system performance degrade over time? How will its performance be affected by different acquisition environments, such as the office environment, public locations, and outdoor?

Generating temporally precise sequences of spikes in response to spatio-temporal patterns of synaptic input is a fundamental process of neural activity: an important example would include coincidence detecting neurons in the primary auditory cortex, that fire in response to synchronous, temporally precise incoming spike trains. This is a non-trivial task however, since typically a neuron receives on the order of 10000 synaptic connections: only a small subset of this input can be considered to convey some meaningful task relevant signal, whilst the rest can be considered noise.

By approximating this random background activity as Gaussian white noise, we investigate how networks of noisy neurons can learn to reliably encode spatio-temporal input spiking patterns as temporally precise output sequences of spikes, specifically through reward modulated spike-timing-dependent plasticity.

Mobile applications only become really useful if combined with cloud-based services. We have observed that the increasingly short time to market may cause serious design flaws in the security architecture. In this talk I will highlight some flaws discovered in the past.

For example, we looked at nine popular mobile messaging and VoIP applications and evaluated their security models with a focus on authentication mechanisms. We find that a majority of the examined applications use the user's phone number as a unique token to identify accounts; they contain vulnerabilities allowing attackers to hijack accounts, spoof sender-IDs or enumerate subscribers. Other examples pertain to (already fixed) problems in cloud-based storage services such as Dropbox.

Innovation in software and product delivery is essential to every company’s success. However, larger companies have particular challenges driving innovation into their products and practices due to their culture, size, and context. In this presentation we look at innovation challenges in largescale software development and delivery organizations. We particularly focus on agile software development methods being used in such organizations and their role in innovation. While there have been many successes in creating innovative software solutions, too often there have been disappointments delivering that software as part of a governed product delivery approach. This presentation looks at the challenges to agile product delivery, and examines the role and characteristics of an agile organization in delivering innovative solutions.

Imaging in ophthalmology has changed faster than our understanding of what these images might mean. The magnitude of the images taken in clinical and research setting and subsequently received for analysis was initially limited by the speed of the slides being developed or by the quality of the polaroids. The last 20 years have seen an unprecedented rise both in the number and the quality of digital images in ophthalmology.

The Ophthalmic Image Analysis Centres (commonly referred to as Reading Centres) now mainly deal with digital images. In broad terms, these come from three sources: screening/clinical settings, epidemiological studies and clinical trials. The requirements for these three are governed by different principles and the time and money involved are vastly different as well. Screening/clinical setting requires fast but reliable approaches some with immediate decision making, others can be slightly delayed, but usually only by days. Epidemiological studies examine thousands of patients and require reliable and reasonably cheap solutions, especially as many patients exhibit no pathology. Clinical trials demand rigorous and time-honoured grading approach with exhaustive quality control approaches.

Both clinicians and researchers have been looking for (semi)automated approaches in order to be able to provide faster and more reliable care to patients with the elimination of at least part of the human factor.

During this seminar we will discuss the clinical and research requirements of image analysis in ophthalmology.

In the last decades a vast empirical and theoretical knowledge about sleep mechanisms has been accumulated. Surprisingly, the function of sleep still remains elusive. Moreover, the question of “why do we sleep?” is now being gradually replaced with a more fundamental one: “what is sleep?” The basic problem with defining sleep is that on one hand it is a state, inasmuch as it is different from other states, such as wakefulness, and on the other hand it is a process, since it is not static, but evolves in time and space in a highly complex manner. There are multiple temporal scales relevant for sleep process – starting from a fast millisecond scale of individual neuron spiking, that changes depending on whether the animal is awake or asleep, and up to a scale of hours or even days, at which the overall amount and architecture of sleep is regulated. With respect to spatial scales sleep also shows astounding complexity – indeed, it is associated with characteristic changes at levels as remote from each other as an individual neuron and the whole brain. Understanding the mechanisms underlying spatio-temporal dynamics of sleep will help us to understand not only what sleep is, but also why it is necessary.