Designing Smart and Secure Communities: from Sensors to Risk-based Reasoning

这项为期五年的计划侧重于确保社区安全的概念和技术，包括人体生物识别传感器和人工智能。这项新兴研究的核心是基于风险的数据概率融合，如人体生物特征(例如，面部或步态)、属性(如步态类型)和背景数据(位置、环境)。该计划的动机是需要实现安全和情报，并平衡它们以解决隐私方面的问题。 该计划有四个主要目标： 1.开发一个基于机器推理的框架，也称为推理，用于系统级决策。在我们的方法中，决策的风险是使用概率推理进行评估的。然后框架还将解决隐私方面的问题，通过使用上下文数据和通过强调行为和属性生物识别来解除个性化，实际上可以大大增强隐私方面的功能。 2.为进一步扩展概率推理理论，开发了一个处理生物特征感知和上下文数据不确定性的多度量推理机。多指标推理引擎以概率因果模型为基础，综合了点概率、区间概率和模糊概率等多种指标，并基于Dempster-Shafer方法进行冲突消解。这将成为基于风险的决策框架的核心。 3.为实际应用提出概念验证解决方案和原型，例如： A)安全的城市、公共交通枢纽、公共活动以及公共建筑。一个例子包括该小组目前正在进行的试点项目，即开发智能避难所出入服务亭，对医疗紧急情况和安全威胁进行基于风险的综合评估。 B)自动过境基础设施的风险评估，包括隐蔽武器探测。一个例子是从一个工业合作伙伴开始的项目，对各种场景进行基于风险的分析，这些场景使用带有和不带有隐藏物体的身体指标。 C)在医疗保健、康复和辅助生活中基于风险的生物识别监测。特别是，拟议的多指标推理引擎将应用于术后患者的监测和风险推断，以及对老年人跌倒风险的预测。 4.实施HQP培训和教育计划，将研究生行业实习以及实践研究和设计部分纳入高年级本科课程，让本科生参与研究，并促进学生的国际合作研究。 从战略角度来看，该计划将有助于(1)推进民用和军用安全应用的计算机辅助系统，(2)开发下一代风险评估工具。(3)在学术和工业领域内协作整合加拿大的研发劳动力。

This five-year Program focuses on the concepts and technologies that make a community secure, including sensors of human biometrics and artificial intelligence. The core of this emerging research is a risk-based probabilistic fusion of data such human biometrics (for example, face or gait), attributes (such as gait type), and contextual data (location, environment). This program is motivated by the need to achieve security and intelligence, and to balance them to address privacy aspects. This Program has four main goals: 1. To develop a framework that is based on machine reasoning, also known as inference, approach to system-level decision-making. The risks of decision-making in our approach are assessed using probabilistic inference. Then framework will also address the privacy aspects, which can actually be greatly enhanced through usage of contextual data and de-personalization by emphasizing on behavioral and attributed biometrics. 2. To further extend the theory of probabilistic inference by developing a Multi-metric Inference Engine that deals with uncertainty of biometric sensory and contextual data. The Multi-metric Inference Engine is based on probabilistic causal models with novel integrated mechanism for multiple metric including point, interval, and fuzzy probabilities, as well as conflict resolution based on Dempster-Shafer approach. This will become the core of the risk-based framework for decision-making. 3. To propose the proof-of-concept solutions and prototypes for the practical applications such as: a) Safe and secure cities, mass-transit hubs, public events, as well as public buildings. An example includes the current pilot project conducted in the group, on developing smart shelter access kiosks with integrated risk-based assessment of medical emergency and security threats. b) Risk assessment in automated border crossing infrastructure including concealed weapon detection. An example is the project started with an industrial partner, on risk-based analysis of various scenarios that use body metrics with and without concealed objects. c) Risk-based monitoring of biometrics in healthcare, rehabilitation, and assisted living. In particular, the proposed Multi-metric Inference Engine will be applied to monitoring and risk inference for post-surgery patients, and prediction of risk of falls in the elderly. 4. To implement an HQP training and educational plan which incorporates industry internships for graduate students as well as practical research and design component in senior undergraduate curriculum, involves undergraduates in research, and promotes student international collaborative research. From a strategic perspective, this Program will contribute to (1) advancing in computer-assisted systems for both civil and military security applications, (2) developing of future generation risk assessment tools. (3) integrating the Canadian R&D workforce in collaboration within the academic and industrial sectors.