A Computational Model of Deception Detection in SMS Scams

短信诈骗中欺骗检测的计算模型

• 批准号: ES/R007764/2
• 负责人: Christopher Street
• 金额: $ 3.76万
• 依托单位: Keele University
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2021
• 资助国家: 英国
• 起止时间: 2021 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=ES%2FR007764%2F2
• 关键词: Computational; Model; Deception; Detection; SMS

This project will develop a model that integrates many aspects of thought in order to explain why people tend to believe that SMS scams are authentic and, in turn, develop a theory-based intervention to reduce the tendency to overbelieve. Because there are many different types of information that are used when making these decisions, and because of the many aspects of thought that come into play, exploring the process by experiments alone would be costly, time-consuming, and limited in terms of determining the mental representations/processes involved. To date, no theory of deception detection has integrated the many aspects of cognition such as attention, goal-planning, and long term memory, all of which are required to decide whether an SMS message is genuine or deceptive. By relying on a well-established and tested theory, the computational model can explore these issues in a cost-effective and relatively time-efficient manner. This project is innovative insofar as it will be the first domain-general model of deception detection.Work package 1 will develop a domain-general computational model in the context of SMS deception detection. The ACT-R cognitive architecture has been extensively tested and developed for over 40 years. It is a theory of the key structures and processes of the human mind that integrates the many aspects of cognition. This project will use the ACT-R framework to implement a computational model of deception detection. The model will be explored to test how people act when (i) their attention is drawn to factors such as an urgent response being required, (ii) the content of the message and the general suspiciousness of the individual are varied, and (iii) an intervention is implemented, such as giving the option to delay responding to an SMS. These factors will be tested with the ACT-R framework in work package 1.Work package 2 will explore the ACT-R model with experimental testing. These studies will mix laboratory control with real-world application. Participants will interact with a mobile phone while their eye movements are recorded. Eye movement data provides precise timing on what information they are drawn to in the SMS scam and the decision of whether reply or delete the text. This precise timing information allows the project to test the predictions of the ACT-R deception detection model.Work package 3 will develop an intervention based on the results of work packages 1 and 2. Consulting with North Yorkshire police and an SMS scam-prevention organisation (The AntiSocial Engineer Ltd.), and in collaboration with a computer scientist, a cell phone-based application will be developed. Although the exact nature of the application will depend on the findings of the first two work packages, based on the current theories in deception detection it is anticipated that the application may scan for key words, encourage users to verify information by searching online, or allow users to press an 'unsure' button, which will hide the SMS message until they can give time to checking the authenticity of the text.This project aligns with the ESRC research priorities of 'ways of being in a digital age' and 'productivity'. Police forces have limited resources to investigate personal SMS scams, while scam-prevention organisations are noting that SMS attacks are on the rise. The outcome of this project - a smartphone application - aims to encourage users to consider the possibility of a scam and to act with a 'healthy scepticism' in the digital world. A reduced risk of engaging with scams will lessen the burden on police forces and aid scam-prevention organisations in delivering evidence-based interventions to potential victims.We will also work with a local child safeguarding board to translate the model into understanding the tendency to overbelieve offenders in the context of social work, where deception is estimated to be involved in 75% of cases in England & Wales where a child is killed.

该项目将开发一个模型，该模型整合了许多方面的思想，以解释为什么人们倾向于相信短信诈骗是真实的，并反过来开发一种基于理论的干预措施，以减少过度相信的倾向。因为在做出这些决定时会用到许多不同类型的信息，而且因为思维的许多方面都在发挥作用，所以仅通过实验来探索这个过程将是昂贵的，耗时的，并且在确定所涉及的心理表征/过程方面受到限制。到目前为止，没有欺骗检测理论已经整合了认知的许多方面，如注意力，目标规划和长期记忆，所有这些都需要决定短信是真实的还是欺骗性的。通过依赖一个成熟的和经过测试的理论，计算模型可以探索这些问题在一个具有成本效益和相对的时间效率的方式。这个项目是创新的，因为它将是第一个域一般模型的欺骗detection.Work包1将开发一个域一般计算模型的背景下，短信欺骗检测。ACT-R认知架构经过了40多年的广泛测试和开发。它是一种关于人类心智的关键结构和过程的理论，整合了认知的许多方面。这个项目将使用ACT-R框架来实现欺骗检测的计算模型。该模型将被探索，以测试人们如何行动时，（一）他们的注意力被提请因素，如需要紧急响应，（二）消息的内容和个人的一般怀疑是不同的，和（三）干预措施的实施，如给予选项，以延迟响应短信。这些因素将在工作包1中的ACT-R框架中进行测试。工作包2将通过实验测试来探索ACT-R模型。这些研究将实验室控制与实际应用相结合。参与者将与移动的手机互动，同时记录他们的眼球运动。眼动数据提供了精确的时间，他们在短信诈骗中被吸引到什么信息，以及是否回复或删除文本的决定。这个精确的时间信息使项目能够测试ACT-R欺骗检测模型的预测。工作包3将根据工作包1和2的结果开发干预措施。咨询北约克郡警方和短信诈骗预防组织（反社会工程师有限公司），并将与一名计算机科学家合作，开发一种基于手机的应用程序。虽然应用程序的确切性质将取决于前两个工作包的发现，但根据当前的欺骗检测理论，预计应用程序可能会扫描关键词，鼓励用户通过在线搜索来验证信息，或允许用户按下“不确定”按钮，这将隐藏短信，直到他们可以给时间来检查文本的真实性。这个项目符合ESRC的研究重点“在数字时代的方式”和“生产力”。警方调查个人短信诈骗的资源有限，而防诈骗组织则注意到短信攻击正在上升。该项目的成果-智能手机应用程序-旨在鼓励用户考虑骗局的可能性，并在数字世界中采取“健康的怀疑态度”。降低参与诈骗的风险将减轻警察部队的负担，并帮助诈骗预防组织向潜在受害者提供基于证据的干预措施。我们还将与当地儿童保护委员会合作，将该模型转化为理解社会工作背景下过度相信犯罪者的倾向，据估计，在英格兰和威尔士，75%的儿童被杀案件都涉及欺骗。