Collaborative Research: CIF: Small: Not All Eggs in One Basket: Authority Distribution for Resilience Against Compromised Nodes in Communication Networks

社会中无处不在的网络通信系统引起了越来越多的安全和隐私问题。特别是，对于收集私人用户信息的应用程序，如健康监测、电子商务和金融服务，以及与关键基础设施相关的应用程序，如交通、智能电网或过程监督管理系统，存在这样的担忧。实际上，对于这些应用程序，单个受到破坏/恶意的代理或网络节点可能引发灾难性事件，例如支付赎金以恢复私有数据、大规模停电、服务中断、数据泄露、财务损失或死亡。该项目旨在探索如何将启动关键任务的权限分配给多个网络节点或用户，而不是单个节点或用户，这是一种有效的方法，可以避免出现单点故障，从而在发生安全漏洞的情况下丢失全部私人数据。此外，拟议的研究还研究了授权分配如何能够在网络被识别为恶意设备或人员之前（最重要的是，在他们能够造成任何伤害之前）赋予网络应对受损设备或人员的能力。该项目研究了超越传统阈值加密的新解决方案，以解决数据访问和生物识别认证的权限分配中的挑战性障碍。具体来说，该项目旨在克服(i)使用生物识别技术的隐私挑战，（ii）需要昂贵的资源作为完全安全的通信渠道，（iii）用户数量的不良扩展行为，以及（iv）需要一个与分布式系统不兼容的中央实体，以解开隐私和存储要求之间的基本权衡。该项目包括三个相互关联的重点。第一个要点研究了在存在（可能不被注意的）受损害的用户或节点的情况下，权限分布可以提供的基本安全和隐私保证。第二个重点研究基于错误控制编码和密码学衍生的新方法，在有限块长度下运行的权限分配的低复杂性编码方案的构建。第三个目标是通过独立于数据统计的深度学习实现权威分配的通用学习框架。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The ubiquity of networked communication systems in society raises ever-growing security and privacy concerns. In particular, such concerns exist for applications that collect private user information, as in health monitoring, e-commerce, and financial services, and applications associated with critical infrastructure as in transportation, smart grids, or process supervisory management systems. Indeed, for these applications, a single compromised/malicious agent or node of the network can provoke catastrophic events such as ransom payments to recover private data, large-scale electricity blackouts, interruptions of service, data breaches, financial losses, or fatalities. This project aims to explore how distributing the authority to launch critical tasks in the hands of multiple, rather than a single, nodes or users of the network can be an efficient means to avoid having a single point of failure and hence losing private data in its entirety in the case of a security breach. Furthermore, the proposed research studies how authority distribution is able to empower networks with the capability of coping with compromised devices or persons before they are identified as malicious and, most importantly, before they are able to do any harm.The project investigates novel solutions beyond traditional threshold cryptography to address challenging roadblocks in authority distribution for data access and biometric authentication. Specifically, the project aims to overcome (i) privacy challenges with the use of biometrics, (ii) the need for costly resources as perfectly secure communication channels, (iii) a poor scaling behavior with the number of users, and (iv) the need for a central entity which is not compatible with distributed systems, to unravel fundamental trade-offs between privacy and storage requirements. The project comprises three interrelated thrusts. The first thrust studies the fundamental security and privacy guarantees that authority distribution can offer in the presence of, potentially unnoticed, compromised users or nodes. The second thrust investigates the construction of low-complexity coding schemes for authority distribution that operate at finite blocklengths, based on novel methods derived from error control coding and cryptography. The third thrust aims to enable a universal learning framework for authority distribution via deep learning that is independent of the data statistics.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.