Adaptive level of detail displays based on machine reasoning methods

网络安全在许多领域日益受到关注，包括民用和国防、零售、工业运营和配电网络。据估计，全球每家公司每年因网络攻击造成的平均损失为770万美元，仅恶意软件每年造成的损失总额就超过5000亿美元。为了保护系统免受新型和日益复杂的攻击，人工网络监控仍然是一项必不可少的任务，但难度很大，而且在认知上具有挑战性。这项研究提出了Focalpoint，一种自适应用户界面（AUI），使运营商能够更有效地监控敏感的网络系统，从而保护数据和关键的网络基础设施。与现有的用户界面相比，Focalpoint跟踪用户交互行为——比如平移、缩放和点击——并自动执行视觉调整，以在不同的任务中唤起适当的注意力水平。这些适应性是通过以下两方面的结合得出的：(1)计算认知模型，允许Focalpoint推断用户的任务和注意力状态；(2)新颖的可扩展机器学习算法，使Focalpoint能够从用户和专家的交互历史中学习。最终的效果是Focalpoint提供了个性化的适应，为每个用户和任务量身定制，将塑造用户的认知状态，以满足他们正在执行的任务的需求。虽然Focalpoint将主要为网络安全运营商开发，但该研究计划将产生通用的AUI原理和技术，即通用的机器学习和推理方法，以及可跨领域转移的视觉界面设计原则。因此，我们设想这项研究将在各种应用领域促进未来的自适应用户界面。**

Cyber-security is an increasing concern in many domains including civil and national defence, retail, industrial operations and power distribution networks. Worldwide annual average losses due to cyber-attacks are estimated at USD $7.7 million per company and total annual malware damages alone exceed USD $500 billion. To protect systems from novel and increasingly sophisticated attacks, manual network monitoring by human operators remains an essential, but difficult and cognitively-challenging task. This research proposes Focalpoint, an adaptive user interface (AUI) that enables operators to monitor sensitive cyber systems more effectively, thereby securing data and critical network infrastructure. In contrast to existing user-interfaces, Focalpoint tracks user interaction behaviors-such as pans, zooms and clicks-and automatically performs visual adaptations to evoke the appropriate attention level across different tasks. These adaptations are derived using a combination of (1) computational cognitive models that allow Focalpoint to infer the user's task and attentional state, and (2) novel scalable machine learning algorithms that enable Focalpoint to learn from user and expert interaction histories. The end effect is that Focalpoint delivers personalized adaptations that are tailored for each individual user and task that will shape the user's cognitive state to the demands of the task that they are performing. Although Focalpoint will be developed primarily for cyber-security operators, this research program will yield general AUI principles and techniques, i.e., general-purpose machine learning and inference methods, and visual interface design principles that are transferable across domains. As such, we envision that this research will foster future adaptive user interfaces in a variety of application areas. **