PhD Studentships in Computer Science at the University of Surrey
The Department of Computer Science at the University of Surrey is offering up to 6 fully funded PhD studentships across its core research areas in cybersecurity and cryptography, distributed and concurrent systems, artificial intelligence and machine learning, computational neuroscience and bioinformatics, amongst others.
Successful applicants will become part of a vibrant PhD community and will benefit from the strong research environment and high international visibility of the Department. Our researchers publish regularly in top conferences and journals and are renowned experts in their fields, developing innovative, practical solutions to real world problems. The Department has strong links with industry and active collaborations with academic institutions worldwide.
Each PhD studentship comes with a stipend of £16,000 per annum plus tuition fees covered for the duration of 3 years for UK/EU candidates. International candidates are welcome to apply but will need to cover the difference between the UK/EU and overseas fees.
The awarded studentships will be allocated to six of the eleven PhD research projects listed below following a competitive selection process. Applicants are welcome to approach mentioned supervisors using the provided contact details to seek more information about the projects.
Project 1. Bayesian statistical modelling of single cell RNA-Seq data from brain tumours
This project will apply statistical methods to the analysis of large scale single cell RNA-Seq data, exploring the use of Bayesian nonparameteric models to incorporate covariates into the analysis. Methods from computational statistics will be applied, for example Markov chain Monte Carlo samplers or variational inference. The project will develop an understanding of cellular heterogeneity in brain tumours in collaboration with researchers at Imperial College London.
Contact: Dr. Tom Thorne.
Project 2. DEVISE: Designing real-life electronic voting systems
This project will design and analyse electronic-voting systems that can be deployed in real-life, by looking at combinations between desirable e-voting guarantees (e.g. privacy, receipt-freeness, collusion resistance, verifiability, accountability, etc) and techniques of executing protocols on untrusted platforms. A secondary aim is to certify the security of these systems using formal-analysis tools.
Contact: Dr. Catalin Dragan.
Project 3. Analysis of string-manipulating programs
Contact: Dr. Taolue Chen
Project 4. Scalable distributed protocols for resilient and trustworthy Blockchains
You will work on distributed agreement protocols that would exploit novel hardware capabilities, such as Remote Direct Memory Access (RDMA), secure execution environments (SGX/CHERY) and non-volatile memory for boosting their performance and failure resilience. You will analyse various aspects of these protocols, such as their security, correctness, and performance, both mathematically and empirically.
Contact: Prof. Gregory Chockler
Project 5. Why do you hate me so? Interpreting and explaining automatic classifications of hate in social media
There are numerous examples where hateful speech has been found online and several engines have been developed, focusing e.g. on racism, misogyny, etc. While they claim to work well on well defined datasets, it is not often clear whether a text is hateful or 'merely offensive'. You will develop new methods to take the context in which a possible hateful utterance is found on social media, and provide interpretations or explanations helping to understand whether something is hateful.
Contact: Prof. Nishanth Sastry
Project 6. Adversarial learning and uncertainty quantification in computer vision
The project will focus on adversarial learning and uncertainty quantification for CV tasks using deep neural networks. The student will investigate adversarial deep learning techniques to train robust and safe neural networks against adversarial inputs. The student will deploy the new methodology in a wide spectrum of computer vision applications including medical image analysis, object detection, pose estimation and object classification.
Contact: Dr. Zhenhua Feng
Project 7. Mapping interpretable models to data, applied to video analysis
We will explore a human-like approach to recognising activities in video. We will use high level models of household activities and build a system that can connect these to video data, by recognising the objects and relationships. Automatic human-activity recognition is crucial for many important monitoring applications: e.g., for residential care monitoring to check on the wellbeing of elderly residents, or to alert on accidents, for urban surveillance, to recognise violence or theft.
Contact: Dr. Frank Guerin
Project 8. Optimisation for cryptanalysis
The security evaluation of symmetric key encryption schemes is a difficult process that involves combinatorial optimisation. Methods such as Mixed Integer Linear Programming, SAT solvers, and more recently, Constraint Programming (CP), have proved to be invaluable tools to assist in this task. This project will extend recent results on differential cryptanalysis using CP to other forms of cryptanalysis, compare different methods, and promote the use of automatic tools.
Contact: Dr. David Gerault
Project 9. Computational modelling of spiking neural network self-organization
Using computer simulations we will study how spiking neural networks can give rise to function, such as classification/prediction of visual signals. This will shine light on how neural circuits in the brain operate and contribute to novel technologies for neurally inspired computing. You will participate in the BioDynaMo project (www.biodynamo.org), visit CERN (Switzerland) for high-performance computing training, and develop open-source code for AI research community.
Contact: Dr. Roman Bauer
Project 10. Uncertainty quantification for trustworthy AI through Bayesian deep learning
Although AI has found its way into our life, we often cannot tell whether AI systems are certain about their decisions. This project will develop Bayesian deep learning techniques centred around optimal transport to quantify uncertainty of AI models, allowing models to understand what they do not know. That way, in case of high uncertainty, we can perform more extensive tests or pass the case to a human in order to avoid potentially wrong results, leading to trustworthy AI.
Contact: Dr. Yunpeng Li
Project 11. One-shot machine learning for automated fraud discovery
This project will explore the latest advances in Human-Like Computing and Third Wave AI for the development of the next-generation machine learning algorithms for automated fraud discovery. We will explore Meta-Interpretive Learning which requires far fewer data samples and its combinations with approaches such as Deep Relational Learning, Federated Learning and Adversarial Machine Learning for the purpose of automated discovery of credit card frauds in real-time.
Contact: Dr. Alireza Tamaddoni Nezhad
How to apply
Apply by the deadline of 21 February 2021. Applications will be assessed on the ongoing basis. We recommend you apply early.