100+ Topics for Computer Science Bachelor's & Master's Thesis

You are looking for a computer science dissertation topic and you keep finding the same words: AI, blockchain, cloud. Almost nobody fails at that stage. Topics fail at the question that comes next: how will you show, at the end, that your solution is any good?

You find a computer science dissertation topic by starting from a field your department actually supervises, narrowing it down to one concrete system and one group of users, and working out before you register what your evaluation data will be: your own measurements, an open dataset, or people who use the thing you have written. By the end of this article you will have three topics to take to a supervisor and a sentence on how you would assess each one.

📌 Key points

• 220 computer science topics across nine fields.
• A working program on its own is not a result.
• The mean SUS score across 500 studies is 68.
• The UK has no AI Act, unlike the EU.
• Ethical approval comes before any data collection.

The topics in this article belong to our overview of dissertation topic suggestions, which covers the other subjects as well.

Want suggestions that fit your degree and your own interests? Enter both into the GoThesis topic generator and the AI will suggest topics for your thesis.

Four steps to the right topic: how to find a computer science dissertation topic

Finding a topic in computer science has one feature no other subject shares: you can produce almost anything you want to study. A dataset, a prototype, a measurement rig, all of it is a question of working hours. That is why computer science topics grow out of hand so easily.

The scale of the work is set by your course. At the University of Sussex an undergraduate Individual Project carries 30 credits at Level 6, the MSc version 60 credits at Level 7. Check your module descriptor first.

The four steps that keep a topic the right size

Each of the four steps below removes one way a topic can fail: no supervisor, too broad a scope, no measurable outcome, and no second opinion. Work through them in order, because the later ones only make sense once the earlier ones are settled. Most students do the fourth step first and the third not at all.

1. Start with the department, not the trend. Look up which research groups your university has and what they supervised recently. Most of those lists are public. A topic with no research group behind it will not find a supervisor, however current it sounds.
2. Narrow a field down to one system. "Artificial intelligence in healthcare" is a field. "A classifier for hospital discharge letters: how well does it identify diagnoses without fine-tuning?" is a topic. The difference is two decisions: which system, and which question.
3. Plan the evaluation before you write code. Most computer science topics fall over here. Sussex expects a working program in its MSc project, but allows a theoretical or experimental study to form a substantial part of it. Either way something has to be measured.
4. Take three topics to the supervisor meeting. Bring one and you get a verdict. Bring three and you get a conversation about the differences, which is the real supervision. Note where the data comes from.

The process for any subject is covered in finding a dissertation topic.

Fields and specialisms in a computer science degree

For dissertation purposes the fields of computer science differ less in content than in where the measurements come from. A proof in theoretical computer science needs nothing but literature. A topic in applied computer science needs an industrial or clinical partner willing to let you near their users. That distinction drives your timetable far more than the technical difficulty does.

The table below sets out what each of the nine fields covers and where its data usually sits. It is deliberately coarse: every field allows a purely theoretical project and almost every field allows a study of your own. The right-hand column names the usual case.

One field is deliberately missing from the list. Anything that asks about business processes, implementation projects and cost rather than about the system itself sits closer to information systems. For that there is a separate list of business informatics topics. If your interest runs the other way, towards the formal underpinnings, the mathematics topic suggestions are worth a look.

220 computer science dissertation topics across nine fields

The 220 suggestions below are sorted by the nine fields above. Treat them as starting points rather than finished research questions: each one still needs a concrete system, a group of users and a measure before it is a topic. "Machine learning in cybersecurity" only becomes a project once it is clear which attack type, which dataset, and measured against what.

Check two things before you take one. First, whether your university supervises the field at all, because a topic without a supervisor will not be approved. Second, whether the technology is still called that. Topics naming a single library or product go stale far faster than topics naming a problem.

Computer science fundamentals: 25 topics

Topics in the fundamentals are the safest choice if you would rather not ask anyone for data. You measure your own system: runtime, memory use, throughput. Setting up the measurement is half the work, and it needs describing well enough that somebody else could rebuild it. The catch: without a baseline to compare against, you end up with a number that means nothing.

• Optimising search algorithms in non-linear data structures
• A comparative analysis of object-oriented programming languages
• Cross-platform application development with modern frameworks
• Automated code reviews and their effect on software quality
• Parallelisation techniques in modern processors
• Evaluating algorithms for graph colouring
• Performance analysis of data compression algorithms
• Evolutionary algorithms applied to optimisation problems
• Synchronisation mechanisms in multithreaded environments
• Designing and implementing a small operating system
• Security aspects of programming languages and compiler construction
• Implementing algorithms on GPU architectures
• Design patterns in functional programming
• Memory management strategies in modern programming languages
• Complexity analysis of algorithms used in cryptography
• Building a framework for automated testing
• Assessing software architectures for scalable applications
• Best practices in API design and development
• Docker in software development and deployment
• Designing and implementing a microservices architecture
• Compiler optimisations for high-level programming languages
• Implementing and assessing debugging techniques
• Real-time operating systems for embedded systems
• Developing a domain-specific emulator
• Continuous integration and continuous deployment in software development

Theoretical computer science: 25 topics

Theoretical projects need no lab and no participants, but they do need stamina. You prove something or you disprove it, and either way the work is paper, pen and a great deal of reading. Such topics assume somebody at your university works in that direction. A proof topic in an applied department rarely finds a supervisor.

• Decidability questions in formal languages
• Complexity classes and the relationships between them
• New algorithms for NP-complete problems
• Applications of game theory in computer science
• Graph isomorphism and its algorithmic challenges
• Quantum computing: foundations and prospects
• Automata and their applications in text processing
• Designing and analysing algorithms for graph problems
• Coding theory and its application in fault-tolerant systems
• Fixed-point theorems in programming language semantics
• Satisfiability problems and modern SAT solver strategies
• Algorithmic game theory and mechanism design
• Petri nets and their application in systems modelling
• Formalising proof systems
• Models of computation beyond Turing completeness
• Category theory in computer science
• The role of recursion theory in computability
• Reduction strategies in complexity theory
• Formal languages and their application in compiler construction
• Randomisation techniques in algorithm design
• Zero-knowledge proofs and their properties
• Formalising concepts of infinite automata
• Fuzzy logic systems in decision-making
• Complexity analysis of distributed algorithms
• Approximation algorithms for NP-hard optimisation problems

Practical computer science: 25 topics

In this field you end up with something that runs, and that is exactly the trap: a working system is not yet a result. Decide early what you will measure to show your solution beats the obvious alternative. Sussex is blunt about the demonstration itself, noting that where programs form part of the project you should expect to be asked to demonstrate a working implementation to your supervisor.

• Assessing RESTful and GraphQL APIs in web applications
• Security strategies for mobile operating systems
• Building real-time data analysis tools
• Best practices for managing large codebases
• Performance analysis of NoSQL databases compared with SQL systems
• Virtual reality in educational applications
• Server-side rendering in web applications: architecture and load times
• Building chatbots with natural language processing
• Implementing security mechanisms in IoT devices
• Drone technology for data collection and analysis
• Developing a face recognition system
• Automated detection and repair of software bugs
• Blockchain technology for secure transaction systems
• A productivity app using gamification elements
• Load-balancing strategies for web servers
• Custom Linux-based operating systems
• Augmented reality for assistance systems in industry
• Efficiency and scalability of cloud storage solutions
• Designing usable data visualisation software
• Algorithms for optimising network routes
• Algorithms for self-healing systems
• Cross-site scripting attacks and how to defend against them
• Developing a forensics tool for digital investigations
• Designing an open-source project management tool
• Implementing and evaluating algorithms for distributed computing

Applied computer science: 25 topics

Applied topics live or die by the partner organisation. A hospital, a charity, a council or a company supplies the case and usually the people for the user test as well. Ask early and ask in writing, because topics in this field fail on access rather than on technique. An applied topic without a partner stays a prototype nobody has assessed.

• Developing a digital health monitoring system
• IT solutions for sustainable agriculture
• Analysing transport systems through simulation and data analysis
• Computing in space flight: navigation and data management
• Digital learning platforms: development and assessment
• IT-supported logistics systems for smart warehousing
• Computational tools in archaeology
• Designing an information system for small and medium-sized firms
• Building a recommender system for e-commerce
• IT security concepts for online banking applications
• Smart home technologies: development and integration
• Analysing social media data for marketing strategies
• Mobile health: app development for patient management
• Virtual reality in psychotherapy
• Developing computer games with educational content
• Computing in forensic analysis
• Implementing ERP systems outside the IT sector
• Information systems in tourism: development and use
• Artificial intelligence in the music industry
• Analysing e-learning strategies at universities
• Smart city: integrating IT systems into urban planning
• Computational approaches to energy management
• IT support for legal data analysis
• Digital art and culture: platforms and archiving
• Developing support systems for people with disabilities

Software engineering: 25 topics

Software engineering carries more empirical projects than any other field. Writing about process models, technical debt or code reviews means collecting statements from developers, and that is good news: practitioners can consent for themselves and most of them enjoy talking about their work. Twelve to twenty completed questionnaires will carry an undergraduate project, though you should agree the number with your supervisor.

• Agile methods and the waterfall model: comparing two process models
• Analysing and improving user experience in software products
• Software metrics and their use in quality assurance
• Developing and assessing continuous-delivery workflows
• Software refactoring: best practices and techniques
• Technical debt and its effect on software projects
• Analysing cross-platform development environments
• Automated test procedures for mobile applications
• Design patterns in cloud-based applications
• AI-assisted software support tools
• Security aspects of the software development process
• Open-source software projects and community management
• The effect of microservices on system architecture
• Software maintenance and evolution: challenges and strategies
• Virtualisation techniques in software development
• Introducing a DevOps culture in an organisation
• Frameworks for software internationalisation
• Assessing architectural styles for enterprise applications
• The effectiveness of bug-tracking systems
• Pair programming and code reviews in practice
• Performance optimisation of high-load web applications
• Project management tools in software development
• Security protocols for web applications
• Artificial intelligence in software testing
• Adaptation mechanisms for legacy software systems

Artificial intelligence and machine learning: 30 topics

Artificial intelligence has shifted further than any other field since 2023, which is why five topics here did not appear in the first version of this list. Two warnings. Compute time is scarce at many universities, so ask about cluster access before you register. And a model you only run against a standard benchmark is not yet a contribution of your own.

• Neural networks for image recognition
• Algorithms for self-learning systems in the financial sector
• Deep learning in medical diagnosis
• Fine-tuning pre-trained language models for a specialist domain
• Machine learning for weather prediction
• Reinforcement learning in autonomous vehicles
• Artificial intelligence in personalised medicine
• Algorithms for pattern recognition in time series
• AI-based optimisation of supply chain management
• Natural language processing for automatic text summarisation
• Artificial intelligence in video surveillance for behaviour analysis
• AI in game development: intelligent NPC behaviour
• Recommender systems for streaming services
• Machine learning in cybersecurity
• Machine learning for stock price prediction
• Algorithms for voice-controlled assistance systems
• Artificial intelligence in agricultural technology for higher yields
• AI methods for analysing and interpreting genomic data
• Machine learning in industrial image processing
• AI-based diagnostic tools in healthcare
• Machine learning for fake news detection
• AI-supported learning platforms
• AI in robotics: developing cooperative robot systems
• Predictive analytics in customer relationship management
• Artificial intelligence for optimising energy grids
• Retrieval-augmented generation: architecture and answer quality
• Evaluating large language models: benchmarks and their limits
• Detecting AI-generated text: methods and error rates
• Quantising neural networks for use on edge devices
• The effect of prompting strategies on language model output

Data science: 25 topics

Data science topics hang on a single question: does the dataset actually exist, and are you allowed to use it? Open portals such as data.gov.uk or Zenodo are the comfortable route. Social network data is no longer comfortable, because the large platforms now charge for their interfaces or have closed them. Check access before you register the topic.

• Frequent pattern mining in large transaction datasets
• Clustering techniques for market segmentation
• Machine learning models for forecasting customer demand
• Deep learning approaches to automated text classification
• Frameworks for stream processing of big data
• Anomaly detection techniques in time series data
• Sentiment analysis in social networks
• Visualisation techniques for complex datasets
• Network analysis methods applied to social media
• Building recommender systems with collaborative filtering
• Predictive analytics for fraud detection
• Analysing spatio-temporal data for environmental monitoring
• Machine learning for optimising logistics networks
• Models for forecasting financial market trends
• Natural language processing for automated news generation
• Optimising algorithms for scalable data processing
• AI techniques for personalising web content
• Techniques for integrating data from multiple sources
• Data management strategies in cloud environments
• Data science in healthcare to improve patient care
• Algorithms for data reduction and compression
• Analysing geodata to support urban planning
• Methods for ensuring data quality
• Text mining to extract information from scientific publications
• Machine learning methods for personalised education

Mathematics for computer science: 20 topics

Mathematical topics are written at a desk and they are more popular with examiners than students expect. A project of this kind shows you understand what sits underneath, rather than which library to call. Five of the twenty points are new and connect to cryptography and machine learning, where the mathematics is currently most visible.

• Elliptic curves and their role in modern encryption
• Modelling and simulating networks with graph theory
• Probability theory in algorithm design
• Numerical methods for differential equations in computer graphics
• Optimisation algorithms in artificial intelligence
• Mathematical models for analysing algorithm performance
• Topology and its application in robotics
• Game theory applied to computer networks
• Analysing encryption algorithms with number theory
• Discrete structures in computer science
• Linear algebra and its application in data processing
• Statistical methods for software quality assurance
• Designing and analysing experiments in computing research
• Mathematical modelling of processes in distributed systems
• Combinatorics in the design of data structures
• Differential privacy: mathematical foundations and practical limits
• Lattice-based cryptography: the mathematics behind the NIST standards
• Numerical stability in neural network training
• Information theory and entropy in data compression
• Markov chains for modelling user behaviour in web applications

Specialist fields and research: 20 topics

Specialist topics are appealing and risky at once. Appealing because there is little literature, so your project contributes something of its own quickly. Risky because little literature also means little to orient yourself by. Take a topic from this field only if your university has a research group working on it.

Two of these points need precise wording today. Post-quantum secure methods have been standardised since 13 August 2024, when NIST published FIPS 203, 204 and 205, among them ML-KEM for key exchange (NIST, August 2024). Anyone writing about quantum computing and encryption should name those standards rather than describe a vague future threat.

• The effect of quantum computing on encryption methods
• Autonomous robot systems for specialised applications
• Ethical aspects of AI systems
• Humanoid robotics and artificial empathy
• Brain-computer interfaces and their applications
• Computational neuroscience in building cognitive models
• Genetic algorithms for problem solving in artificial intelligence
• Advanced systems for fog computing
• Edge computing in IoT environments
• Cyber-physical systems and the industrial internet
• Mixed-reality applications for education and training
• Nanotechnology in computing
• Migrating existing systems to post-quantum secure methods
• Bioinformatics: methods for analysing genetic sequences
• Computing for sustainable energy
• Conformity requirements for high-risk AI systems under the EU AI Act
• Energy use in training large models: measurement and reduction
• Reproducibility in computing research: artefacts and badges
• Web accessibility under the WCAG: testing methods and tools
• MLOps: operating and monitoring models in production

Which computer science topics need a study of your own

Which computer science topics need a study of your own is given away by their wording: anything asking about acceptance, usability or use in practice requires statements from people. Everything else comes from your own measurements or from the literature. Settle that question before you register.

The table below maps the commonest wordings to what they demand. Every row with "your own survey" or "your own user test" on the right means roughly four extra weeks for the questionnaire, consent and recruitment.

For those rows, decide first whether you want to count or to understand. Counting needs a standardised questionnaire and enough participants. Understanding gets a long way on eight good interviews. The difference is set out in our article on the online survey as a research method.

One point goes wrong with some regularity: representativeness. A survey of 40 coursemates says nothing about developers in the UK, and it does not need to. Describe it correctly and it stands. When a survey can carry that claim is explained in our article on representative online surveys.

Evaluating your system: the part most computer science dissertations are missing

You evaluate your system by deciding, before the first line of code, which goal you are testing, with which question and with which measure, and then setting the result against a baseline or against the judgement of real users. Without that plan, all you can report at the end is that the software runs.

Computer science is the subject where students most often build something and least often think beforehand about how to assess it. The prototype runs, the repository is full, submission is two weeks away, and the part that makes the project academic work is missing. Evaluation therefore belongs in the choice of topic, not in the closing chapter.

Set the evaluation plan before the first line of code

An evaluation is planned, not bolted on. One established format for it is the GQM plan: goal, question, metric. You state the goal first, then the question, then the measure that answers it. The Software Design and Quality group at KIT makes it mandatory for master's projects with an empirical validation and recommends it for bachelor's projects.

The same group draws a floor under it: in a master's project the evaluation has to go beyond purely functional tests. Unit tests show that your code does what you programmed. They do not show that your solution is better than what existed before. Your own course handbook is what binds you, but the principle holds across subjects.

Ethical approval comes before data collection, not after

Any project involving human participants needs ethical approval first, and UK universities state this without much softening. UWE Bristol writes that approval must be obtained "before any primary data collection for the project begins", and that it applies to research by students at both undergraduate and postgraduate level wherever human participants are involved. Build the application into your timetable, because a committee does not turn things round overnight. What your own institution requires is set by its research ethics committee, not by this article.

Measuring usability: the System Usability Scale

As soon as people are meant to use your system, you need a measure of their judgement. The most widely used questionnaire for that is the System Usability Scale, or SUS: ten statements, a five-point response scale, and a score between 0 and 100 at the end. The value for a dissertation lies in the benchmark. Jeff Sauro analysed 500 studies in 2011 and writes: "The average SUS score from all 500 studies is a 68."

The benchmark gives your result an interpretation no home-made satisfaction question can offer. A SUS score of 74 is above average, one of 55 is not. Mind the limit, though: SUS measures the subjective judgement of usability, not task success. How long a task takes and how often it fails are things you measure alongside it.

Reproducibility: what stays checkable after you submit

Reproducibility is a quality marker in its own right in computing, and a useful line of defence in a viva. The Association for Computing Machinery, or ACM, draws a distinction in its policy on artefact review and badging (version 1.1, 24 August 2020): "Results Reproduced" means another team obtained the results using the authors' artefacts, while "Results Replicated" means another team obtained them without those artefacts.

For your own project that boils down to a short list. Deposit source code, configuration, dataset and analysis scripts so that your supervisor could recompute your numbers. Give the versions of the libraries you used, not just their names. And record which runs you discarded. A project whose numbers can be recomputed is harder to attack than one with prettier results.

For a survey or a user test you need a tool to build the questionnaire and analyse the responses. empirio.ai, an online survey tool from Germany, allows 50 responses a month on its free tier and hosts in the EU, with a data processing agreement (as of October 2026). For a SUS study with 30 participants that is enough. Several hundred responses need a paid tier, and that belongs in the picture too.

Topics for an MSc dissertation, a final year project and a taught coursework project

Topics for an MSc dissertation, a final year project and a smaller coursework project come from the same list, cut to different sizes: the coursework project takes a single measure, the final year project a solution plus an assessment of it, the MSc dissertation a comparison of several methods. What matters is not the suggestion but how far you open it up.

The cases below come up most often in supervision meetings. The same principle holds for all of them: the shorter the piece of work, the narrower the question and the smaller the evaluation.

Scaling a topic up to an MSc dissertation

An MSc dissertation in computer science usually differs from an undergraduate project not by more code but by more comparison. "Building a recommender system for e-commerce" becomes, at master's level, a comparison of three methods on the same dataset, with significance testing and a discussion of the limits. The scope grows where the argument grows, not where the line count grows. Sussex caps the written output for its informatics MSc dissertations at 12,000 words, excluding the abstract, footnotes, tables, bibliography and appendices.

Cutting a topic down to a final year project

For a final year project you take a suggestion from the list and strike out everything comparative. "Agile methods and the waterfall model: comparing two process models" becomes "Retrospectives in one Scrum team: how they run and what the team gets from them". A final year project needs no benchmark; it needs a clean description and one claim you can support. Further examples sit in our list of coursework and term paper topics.

Computer science or information systems: where a topic belongs

Computer science and information systems overlap heavily at many universities, but the examined question differs. Computer science asks how a system works and whether it works better than another. Information systems asks whether a system is of use to an organisation and what introducing it costs. In practice: if the measure in your topic comes from the system, such as runtime or precision, it is computer science. If it comes from the organisation, it is information systems.

Once the topic is settled, the research question is the next step. A research question is not the same thing as a topic: it has to be answerable, and answerable with the data you will actually get. The difference between a question, a thesis statement and a hypothesis is explained in our article on research questions and hypotheses.

Common mistakes when choosing a computer science topic

Four mistakes come up especially often when choosing a computer science topic: a technology name used as a topic, a prototype with no baseline, a legal framework taken from the wrong country, and no access to the data. All four share a root: the topic was thought out from the technology rather than from the question.

All four can be cured before you register. Afterwards they cannot, because topic and title are generally fixed by then and changing them means registering again.

The topic is a technology, not a question

"Blockchain in logistics" is a field, not a dissertation. The mistake happens because technology names are the familiar vocabulary from lectures and job adverts. The result is a project that spends thirty pages explaining what the technology is and five on what you did with it. The way out is mechanical: put a system, a group of users and a measure behind your topic. If it is still too big, cut it in half.

The prototype runs but there is nothing to show

The commonest write-off in computer science projects: the system works, submission is close, and no baseline exists. The results chapter then states that the application was "successfully implemented", and that earns no marks. Before the first commit, fix what you are comparing against: an existing solution, a simple reference method, or the state of things before your change.

Assuming the EU AI Act applies in the UK

The commonest legal error in UK computer science projects is lifting the EU framework from a German or Irish source. The UK has not adopted an EU-style AI Act. Its published approach is principles-based and delivered through existing regulators: the government's white paper A pro-innovation approach to AI regulation, presented to Parliament on 29 March 2023, states "Initially, we do not intend to introduce new legislation" and that "The principles will be issued on a non-statutory basis and implemented by existing regulators".

For your topic that means two things. If your system is intended for the EU market, or its output will be used there, the EU regulation is in scope and you should say so. If it stays in the UK, describe the principles-based framework and the regulator that covers your sector, with a date. UK GDPR still applies to personal data either way. This article is not legal advice.

No access to the data you intend to work with

A topic on sentiment analysis on a social platform reads well and often ends at a paywall. Interfaces that were open in 2020 now cost money or have closed. Company data has the same problem: a verbal yes during a placement is not clearance from the legal department. For official statistics data.gov.uk and the ONS are the first stops; the ONS labour market series put employment in the information and communication industry at 1,731,000 for the second quarter of 2026. Download the dataset or get the written yes before you register.

Checklist before you register

• System, users and measure appear in the working title.
• The baseline for the evaluation is named.
• The dataset is in hand or the permission is in writing.
• Ethical approval is applied for, not assumed.
• Participants for a user test are reachable.
• Compute time and tooling are sorted.
• A research group at your university covers the field.

A computer science topic is not good because the technology is exciting. It is good because you can measure something at the end.

Conclusion

Take three of the 220 suggestions that fit a research group at your own university, then choose between them on the evaluation rather than on the interest. Topics from theoretical computer science and mathematics always work, because you need nobody's permission for data. Anything with a user test or a survey needs four weeks of lead time and enough participants.

What to do next

• Want to know how long the dissertation should be and which chapters belong in it? Dissertation structure and guide
• Need a questionnaire for your user test? How to create a questionnaire
• Looking for a ready-made template to start from? Survey and questionnaire templates
• Got your topic and now need a research question? Formulate research questions with the GoThesis generator

Does your topic need a user survey?

Build the questionnaire in a few minutes, share the link with your test group and have the responses analysed straight away.

Create a survey for your dissertation