FinFraudSIM: Financial Fraud Simulative Analytic Research Platform
Project team
Summary
FinFraudSIM aimed to address the problem of online financial fraud. Fraud is the most common crime type in the UK. Extrapolating from the Crime Survey for England and Wales there were over 3.8 million fraud incidents in the year ending in September 2024 (1). In 2023, 80% of fraud cases were deemed to be cyber offences (2). Financial fraud can cause great harm to individuals, businesses and other organisations, such as charities, as well as to the public sector and the whole of society. Such harms include financial loss, emotional distress, and loss of trust in society. According to the UK Finance Annual Fraud Report, in 2023 alone over £1.1 billion was stolen through fraud (3). Despite considerable efforts by the police, Government, financial institutions and other stakeholders, fraud is still on the rise with a 19% increase in the number of incidents in 2024 compared to the year before (1).
The complex and dynamic nature of financial fraud, particularly online fraud, demands innovative intervention strategies. While some technologies, like behavioural biometrics and device analysis, have proven effective in early fraud detection, a comprehensive, interdisciplinary approach is crucial to ensure maximum protection for end users, especially as reimbursement laws evolve. Existing research related to online financial fraud has been often siloed with social scientific and technological perspective rarely converging. This fragmented approach fails to fully address the sociotechnical complexities of fraud. Moreover, current detection and prevention tools are often ad hoc developments, focusing primarily on financial organisations rather than a broader societal context. This gap in comprehensive understanding and integration limits the effectiveness of current interventions.
To address this gap, FinFraudSIM applied an interdisciplinary approach to build on knowledge bases in relevant areas, including criminology, AI and ML, human-centred computing, and Science and Technology Studies. The project utilised the criminological approach of Crime Script Analysis (CSA), and AI and ML-based technologies to design a platform which can serve as a decision-support tool for professionals working in preventing online financial fraud.
Crime Script Analysis is a methodology used in criminology to understand the processes and sequences of actions involved in committing a crime (4, 5). It breaks down criminal activities into distinct steps or stages, much like a script for a play, allowing researchers, law enforcement, and policymakers to analyse the crime in a detailed and structured manner. The method can be used to extract known cases of crimes, such as online financial fraud, and used to prevent the crime by disrupting any one of these steps. Crime script analysis has been used to explore financial crime as well as crimes online (6, 7). However, so far research on examining the capabilities of AI/ML in crime script analysis has been limited (8).
This project served as a foundational proof-of-concept study to understand what we need to know and have in place for a larger, more ambitious funding proposal. By investigating the integration of ‘Crime Script Analysis’ with AI, we laid the groundwork for proposals to funding bodies such as UKRI, ESRC, Innovate UK, and RAI-UK. The insights gained from this project have been critical in shaping our long-term research agenda.
Our long-term objective is to develop an ambitious and innovative workbench named ‘Financial Fraud Simulative Analytic Research Platform (FinFraudSIM)’. This platform will serve as an experimental environment that models and visualises interactions between fraudsters and victims, simulating the effects of various interventions. The goal is to identify effective intervention points for mitigating financial fraud in a controlled yet dynamic environment. The uniqueness of FinFraudSIM lies in its genuinely interdisciplinary design, combining cutting-edge approaches from criminology, data science, behavioural analytics, and human-centred computing. This platform will conceptualise financial fraud as an activity occurring within a sociotechnical digital assemblage, drawing on expertise from both technical and social-scientific fields.
References
(1) Office for National Statistics (2025) Crime in England and Wales: year ending September 2024. https://www.ons.gov.uk/peoplepopulationandcommunity/crimeandjustice/bulletins/crimeinenglandandwales/yearendingseptember2024
(2) NECC (2024) National Economic Crime Centre Annual Report, 2023-2024., pp.4. https://www.nationalcrimeagency.gov.uk/who-we-are/publications/730-national-economic-crime-centre-annual-report-2023-2024/file
(3) UK Finance (2024) Annual Fraud Report 2024., pp.11. https://www.ukfinance.org.uk/policy-and-guidance/reports-and-publications/annual-fraud-report-2024
(4) Cornish, D. (1994). The Procedural Analysis of Offending and its Relevance for Situational Prevention. In Clarke, R.V. (Ed.) Crime prevention studies, Volume 3 (pp. 151–196). Criminal Justice Press. https://popcenter.asu.edu/sites/g/files/litvpz3631/files/problems/stolen_goods/PDFs/Cornish1994.pdf
(5) Ekblom, P., & Gill, M. (2016). Rewriting the Script: Cross-Disciplinary Exploration and Conceptual Consolidation of the Procedural Analysis of Crime. European Journal on Criminal Policy and Research, 22(2), 319–339. https://doi.org/10.1007/s10610-015-9291-9
(6) Hutchings, A., & Holt, T. J. (2015). A Crime Script Analysis of the Online Stolen Data Market: Table 1. British Journal of Criminology, 55(3), 596–614. https://doi.org/10.1093/bjc/azu106
(7) Van Nguyen, T. (2022). The modus operandi of transnational computer fraud: a crime script analysis in Vietnam. Trends in Organized Crime, 25(2), 226–247. https://doi.org/10.1007/s12117-021-09422-1
(8) Lwin Tun, Z., & Birks, D. (2023). Supporting crime script analyses of scams with natural language processing. Crime Science, 12(1), 1. https://doi.org/10.1186/s40163-022-00177-w
Objectives
The FINFRAUDSIM proof-of-concept project was structured around four interrelated objectives, each supported by targeted activities designed to lay the groundwork for a future large-scale research programme.
Objective 1: Literature Review: Exploring the online financial fraud landscape and identifying gaps in existing interventions
The aim of this objective was to analyse the current landscape of online financial fraud, focusing on identifying the limitations and gaps in existing intervention strategies, paying particular attention to Crime Script Analysis.
Objective 2: Development of a non-functional Mock-up of FinFraudSIM Platform Using Crime Script Analysis
The aim of this objective was, firstly, to identify tasks related to the development of the FinFraudSIM platform, based on Crime Script Analysis, where AI can be incorporated. Secondly, to determine what data is needed to perform these tasks (e.g., relevant data sources), the associated data challenges, their practical consequences, and the limitations of using such data. Thirdly, to develop a mock-up of the FinFraudSIM platform.
Objective 3: Seeking Feedback on the Mock-up from End-Users for Refinement
NOTE: This objective was changed.
The aim of this objective was to implement a meaningful engagement and knowledge exchange strategy through co-design processes by organising two exploratory interviews with relevant end users from the private sector (e.g., from the financial industry). The two exploratory interviews with experts were to form a basis upon which future focus group meetings and interactive workshops in the bigger-scale project would be built.
We have contacted relevant experts and have scheduled two interviews to take place in the second half of September 2025. In the interviews the mock-up will serve as a stimulus for discussion around usability, real-world relevance, and potential applications. Insights will inform refinements to the platform’s design and underlined the need for co-created, practice-oriented fraud prevention tools.
Objective 4: Outlining a Large-Scale Research Grant for Further Development of FinFraudSIM
The aim of this objective was to begin preparing a large-scale bid to potential funders such as ESRC or EPSRC in order to further develop FinFraudSIM. This included identifying a suitable grant to target and establishing potential collaborations with relevant organizations.
Activities
1. Review of fraud landscape in England and Wales
Corresponding Objective: O1
We conducted a review of the fraud landscape with reference to the main categories of cyber-enabled fraud estimated by the Crime Survey for England and Wales (CSEW). The review considered how fraud is defined and classified in the CSEW and under the police recording rules. To determine which fraud types to prioritise in the analysis, we assessed several key factors, including prevalence and relevance in terms of scale and impact; data availability; suitability for CSA from an offender-victim interaction perspective; and offending and victimisation characteristics. The review drew on legislative and regulatory updates, CSEW estimates, literature review, and search for indictment files. Findings are summarised in an internal report.
2. Review of CSA in online financial fraud
Corresponding Objective: O1
We developed a PRISMA-guided scoping review protocol and conducted a systematic search across peer-reviewed academic publications and grey literature. The aim was to scope and synthesise existing evidence on the use of Crime Script Analysis (CSA) in the context of online/digital fraud, and to identify opportunities for integrating CSA with AI/ML methods.
This involved mapping the online financial fraud landscape, highlighting current intervention strategies, identifying gaps and limitations, and exploring data challenges and ethical considerations to set a clear research agenda.
The review spanned criminology, social policy, behavioural science, and computer science, ensuring an interdisciplinary perspective. We extracted script stages, actors, data sources, and any tested or proposed prevention measures. We are also producing a narrative synthesis of the findings.
3. Review of interventions in online romance fraud
Corresponding Objective: O1
We conducted a non-systematic review of interventions in online romance fraud. Through systematic Scopus inquiries and additional Google searches we identified 23 papers focussing on, or having a more substantial section on, interventions targeting online romance fraud. We extracted types and methods of interventions, as well as information available on effectiveness and impact.
4. A ‘model’ CSA
Corresponding Objective: O2
The aim of creating a ‘model’ of CSA was to demonstrate the methodology of conducting CSA in the context of fraud, with a view to support the interdisciplinary collaboration on the development of FinFraudSIM. We developed a step-by-step guide that outlines how to perform CSA, beginning from document review, scene-actor-role-resource identification, and how to segmentate scenes based on Cornish (1994) general script framework. We then demonstrated this by applying CSA to the indictment of a romance fraud offending group, where we manually coded relevant sections into scripts scenes, identified actors and roles, standardised actions and tools, and structured outputs into tabular format. This demonstrated the methodological requirement of CSA from a social science perspective, and provided a concrete annotated example to inform the design of FinFraudSIM.
5. FinFraudSIM Design requirements
Corresponding Objective: O2
The FinFraudSIM platform design requirements, identified through literature review and iterative design sessions, envision a comprehensive system capable of (a) collecting unstructured crime-relevant data from diverse sources such as police reports, court cases, and news articles, (b) processing this data using AI-driven analysis to identify common Crime Script Analysis (CSA) elements and model them into structured crime scripts with defined stages and opportunities, (c) enabling iterative model refinement through human-AI collaborative interfaces similar to modern generative AI systems, and (d) visualizing the resulting models through accessible, interactive web-based platforms or reports.
6. A mock-up of FinFraudSIM
Corresponding Objective: O2
Corresponding Objective: O3
Using insights from the literature review, we identified the main tasks related to developing an experimental simulation platform based on CSA. We mapped out relevant data sources, assessed practical challenges (e.g., data availability, quality, and ethical use), and considered the role of AI in modelling fraud processes. Building on this foundation, we developed both a static, non-functional mock-up of the FinFraudSIM platform in the form of a flowchart illustrating the platform process, and a comprehensive interactive crime script analysis system that serves as a targeted proof-of-concept demonstrating the critical visualization and stakeholder engagement components (requirement (d)) of the FinFraudSIM vision.
The interactive system, built using standard web technologies and populated with verified data from official UK sources, showcases how complex crime script frameworks can be transformed into intuitive, interactive interfaces that effectively communicate research findings to diverse professional audiences. This web-based demonstration illustrates how financial fraud scenarios can be visualised, broken down into crime script stages, and tested against different interventions, while the system's clean design, comprehensive content organization, and multi-stakeholder functionality provide a concrete example of how FinFraudSIM's analytical outputs could be presented, validating the platform's potential for real-world application across law enforcement, financial institutions, and policy-making organizations.
The developed mock-ups served as key design tools for communicating the project's vision and facilitating engagement with stakeholders. The proof-of-concept establishes the foundation for stakeholder engagement by demonstrating the end-user experience and facilitating expert feedback collection, which will inform the development of the complete FinFraudSIM platform's data collection, processing, and AI-collaborative refinement capabilities. The aim is to ensure that the developed FinFraudSIM system effectively bridges academic research and practical application, providing stakeholders with a tangible demonstration of how automated crime script analysis could enhance fraud prevention and investigation efforts.
7. ‘Laying the groundwork’ for a large-scale application
Corresponding Objective: O4
Drawing on insights from the literature review, mock-up development, and stakeholder engagement, we are developing plans for a future large-scale grant proposal. We established contact with potential partners (from academia and industry), and identified routes to reaching other potential stakeholders (in public sector and industry). The proposal framework is set to entail the next phase of research which is to scale up the proof-of-concept into a fully functional, interdisciplinary platform.
8. Dissemination and knowledge exchange
Corresponding Objectives: O1, O2
Project outputs in progress include two academic articles: one drawing from the literature review, and another on using AI/ML to enhance Crime Script Analysis. Findings were also presented at the SPRITE+ 2025 Showcase (08/09/2025).
Outputs
1. A mock-up design of FinFraudSIM:
To facilitate the understanding of end user needs and to develop solutions that meet these needs, interactive prototypes and mock-ups were developed to illustrate the different use cases for FinFraudSIM. These comprehensive crime script analysis interactive systems based on Tompson and Chainey's research framework were built using standard web technologies and populated with verified data from official UK sources such as Action Fraud, UK Finance, and the ONS, serves as both a functional demonstration of the FinFraudSIM concept and a tool for stakeholder engagement.
The system's integration of real-world crime patterns, evidence-based intervention strategies, and UK-specific romance fraud statistics ensures that the interventions created are not only practical and user-friendly but also grounded in established academic research and capable of being effectively implemented in real-world scenarios by law enforcement, financial institutions, and fraud prevention organizations. This objective emphasises the importance of stakeholder engagement in the design and development of FinFraudSIM, with the interactive crime script system serving as a concrete example for obtaining expert feedback from diverse end users (including investigators, researchers, victim support specialists, and policy makers) on both the mock-up interface and the underlying analytical framework, enabling further refinement based on professional expertise and practical implementation requirements
2. Articles in progress
-
Article 1: Crime Script Analysis of Online Fraud: A Scoping Review - We are currently writing up a systematic literature review on the use of Crime Script Analysis in online financial fraud, set for submission to the journal Computer in Human Behaviour Report.
-
Article 2: [Name of the article to be decided] - This article will describe the development and potential applications of the FinFraudSIM platform. [intended journal to be confirmed].
In addition to the two FinFraudSIM articles above, Dr Xiaochun Cheng (Co-I at Swansea University) led and supervised a group of master’s students in computer science to write and submit three academic articles connected to the project’s theme (these have not yet been accepted for publication).
3. Presenting the project at SPRITE+ Showcase, 8 September 2025 (around 45 participants in attendance)
Impact
Our project delivered several key insights on online fraud and the use of CSA. Findings from our scoping review showed that CSA is increasingly used in fraud research, but coverage is often partial that maps only segments of the script. They also highlighted the predominant use of case file and online forum data, with only few studies triangulating offender, victim, and crime preventors perspectives. Evaluations of interventions are found to be rare. The evidence table we produced gives stakeholders a common framework to align modus operandi of fraud and prevention through a holistic view of the crime commission process.
In building the CSA model that details the methodological steps, we demonstrated that CSA can be systematically applied to legal documents such as indictments. In doing so, it shows the practical implication of adopting CSA to monitor modus operandi, support cross-case learning, and offers a structure for targeted intervention analysis for investigators. Further, we also highlighted the limitation of such data which only allowed certain scenes of the crime script could be reliably extracted due to limited narrative details typical in court documents. This clarified requirements for the development of an AI-assisted CSA tools, by highlighting the required data inputs and data sharing priorities, annotation structure, linguistic cues, and the systematic structure (universal script) that processes the data. This will help shaping a more accurate and actionable evidence base for disrupting online financial fraud.
Future work
1. We are planning to complete two distinct academic papers which are in-progress at the time of submitting the report (see above for details).
2. We are planning to identify requirements for developing a functional prototype of FinFraudSIM. Such requirements include, and are not limited to, data access, and partnerships with industry/public sector, as well as to identify a suitable UK grant.
3. We are planning to submit a larger-scale grant application to a UK funder in order to be able to develop a functional version of FinFraudSIM.