QoIA-Aware Attack Resilient Database Systems

该项目将构建一个QoIA感知的抗攻击数据库系统框架，称为Linba。随着人们面临越来越多的网络安全威胁，计算系统在面对攻击时提供有效服务的可信度已经成为一个比以往任何时候都更加关键的问题。信息质量保证(QOIA)服务是与特定的可信性级别相关联的服务。从终端用户的角度来看，可信计算的目标是让人们在面临攻击的情况下也能获得他们已经订阅的服务。然而，由于现有可信系统在提供(持续)定量可信度保证方面的能力非常有限，现有的可信系统大多不能提供服务，本研究的目的是构建一种新的可信计算基础设施，能够在面对攻击和成本约束时继续提供服务。基于一种新的形式(服务)可信度模型，Linbadelver通过近乎最优的神经模糊复合QOIA自适应来实现多层次、差异化、量化的QOIA服务，其中LINBA智能地自适应环境变化，并以优化的方式完成QOIA与成本之间的权衡。Linba的成功开发将使现有的可信数据库系统具备提供QoIA服务的能力(以具有成本效益的方式)，并为开发其他类型的能够提供QoIA服务的可信计算系统提供非常有价值的提示。

This project will build a QoIA-aware attack resistant databasesystem framework, call Linba. The trustworthiness of a computingsystem in delivering valid services in face of attacks hasbecome a more critical concern than ever as people are experiencingincreased cyber security threats. A Quality of InformationAssurance (QoIA) service is a service associated with a specific levelof trustworthiness. From the viewpoint of end users, the goal of trustedcomputing is to enable people to get the QoIA services that theyhave subscribed for even in face of attacks. However, (most) existingtrusted systems cannot deliver QoIA services since they have very limitedability in providing (sustained) quantitative trustworthiness guarantees.The objective of this research is to build a new trusted computinginfrastructure that is able to continue delivering QoIA servicesin face of attacks and cost constraints. Based on a novel formal(service) trustworthiness model, Linbadelivers multilevel, differential, quantitative QoIA servicesthrough near optimal neuro-fuzzy composite QoIA adaptations whereLinba intelligently adapts itself to environment changes andQoIA-cost tradeoffs are done in an optimized way. Successfuldevelopment of Linba will arm existing trusted database systemswith the ability to deliver QoIA services (in a cost-effectiveway) and provide very valuable hints on developing a variety of othertypes of trusted computing systems that can deliver QoIA services.The cost-effectiveness of Linba will be evaluated throughsimulation or prototyping.