Safety Advancing Federated Estimation of Risk using AI (SAFER AI)

使用人工智能推进安全联合风险估计 (SAFER AI)

• 批准号: 10093091
• 金额: $ 148.3万
• 依托单位: THE DATA ANALYSIS BUREAU LTD
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2024
• 资助国家: 英国
• 起止时间: 2024 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=10093091
• 关键词: Safety; Advancing; Federated; Estimation; Risk

Data privacy and security is a key concern for the adoption of Artificial Intelligence/Machine Learning with IoT (Internet of Things), particularly where that data is personal or security sensitive. The number of IoT-enabled devices and machines is growing exponentially, estimated to reach 26Bn globally by 2030\. Creating platforms optimised to maximise private and secure machine learning at the edge in distributed systems that can be trusted is an urgent priority.The consortium's mission is to help solve the challenge of implementing trustworthy AI-in-IoT. This will be achieved by accelerating the development of a federated, secure, privacy-preserving, and auditable AI-for-IoT platform optimised for machine learning in IoT and edge systems. OctaiPipe is a first-of-its-kind innovation that combines privacy-preserving machine learning technology, cyber security, continuous collaborative learning and AI lifecycle management. This will allow IoT-enabled businesses to build, deploy, and manage machine learning software that guarantees the privacy and security of device data and its use, allowing the user to have a high degree of trust in the AI solutions embedded in them.Many organisations already collect high-level operational and HSE incident data intelligence through various Industrial IoT devices and cameras to successfully predict safety incidents. However, analytics based on this is not meaningfully actionable to drive changes that preventatively reduce risks.For predictions to be meaningfully actionable, they must be made at a sufficient level of granularity within the workgroup. Technology now exists to monitor such events at granularity---enabling the build of models to predict and forecast events in the future at sufficient granularity---enabling game-changing preventative impact at the workgroup level.Fortunately, HSE critical events are rare within single sites or even organisations---meaning insufficient data exists to employ ML models capable of predicting when and why H&S incidents might occur so they can be prevented. However, progress towards an AI-enabled solution is impeded by:a) a requirement for more observations than one organisation can generate alone, so it is imperative to share data, andb) barriers to sharing data across organisations that, until now, have not been overcome.Federated Learning solves this. The project will enable organisations to combine data to facilitate actionable incident predictions for small work groups.This project aims to address vulnerabilities in Machine Learning for IoT with a specific focus on FL and addresses the fundamental challenges of socially responsible AI adoption into society.

数据隐私和安全是采用人工智能/机器学习与物联网（物联网）的一个关键问题，特别是在数据是个人或安全敏感的情况下。支持物联网的设备和机器的数量呈指数级增长，预计到 2030 年全球将达到 260 亿台\。当务之急是创建优化平台，以最大限度地提高可信分布式系统边缘的私有和安全机器学习。该联盟的使命是帮助解决实施值得信赖的人工智能物联网的挑战。这将通过加速开发联合、安全、隐私保护和可审计的物联网人工智能平台来实现，该平台针对物联网和边缘系统中的机器学习进行了优化。 OctaiPipe 是一项首创创新，结合了隐私保护机器学习技术、网络安全、持续协作学习和人工智能生命周期管理。这将使支持物联网的企业能够构建、部署和管理机器学习软件，保证设备数据及其使用的隐私和安全，从而使用户能够高度信任其中嵌入的人工智能解决方案。许多组织已经通过各种工业物联网设备和摄像头收集高级运营和 HSE 事件数据情报，以成功预测安全事件。然而，基于此的分析对于推动预防性降低风险的变革并没有有意义的可操作性。为了使预测具有有意义的可操作性，必须在工作组内以足够的粒度级别进行预测。现在的技术可以在粒度上监控此类事件，从而能够构建模型来以足够的粒度预测和预测未来的事件，从而在工作组级别上实现改变游戏规则的预防性影响。幸运的是，HSE 关键事件在内部很少见。单个站点甚至组织——这意味着没有足够的数据来使用能够预测健康与安全事件何时以及为何发生的机器学习模型，从而可以预防它们。然而，人工智能解决方案的进展受到以下因素的阻碍：a) 需要比一个组织单独生成的观察结果更多的数据，因此共享数据势在必行；b) 跨组织共享数据存在障碍，而迄今为止，这些障碍还没有解决联邦学习解决了这个问题。该项目将使组织能够整合数据，以促进小型工作组进行可行的事件预测。该项目旨在解决物联网机器学习中的漏洞，特别关注 FL，并解决社会采用具有社会责任感的人工智能所面临的基本挑战。