SGER: Monitoring Off-the-shelf Components

这笔用于探索性研究的小额拨款调查了一种很有前途的新方法，用于检查由组件组成的软件的正确执行。开发新的反应性软件的一个有吸引力的方法是使用现成的？不完全满足或可能不完全满足所需要求的第三方组件。为了安全地使用这些组件，使用运行时监视器来检测违反要求的行为。这样的需求通常由安全部分和活跃性部分组成，前者的运行时监控得到了很好的研究，而后者的运行时监控却难以捉摸。通常情况下，运行时监控器提取的安全属性要么过近似，要么欠近似，因此往往具有很高的不精确度。粗略地说，这种方法偶尔会抛硬币来决定是否放弃达到活跃度要求。这种方法非常准确，并且具有优雅的降级特性：活性特性不满足的时间越长，它就越有可能被监视器拒绝。该项目还研究了评估监测方法准确性的新方法，以比较各种算法。最后，监视器通常是被动的--他们只观察计算，而不参与计算？但这项工作是探索的？活跃的？与组件协作以生成满足要求的计算的监视器。

This small grant for exploratory research investigates a promising new approach to checking the proper execution of software composed of components. An attractive method for developing new reactive software is to use ?off-the-shelf? third-party components that do not, or may not, exactly satisfy the desired requirements. To safely use such components, the use of run-time monitors is used to detect behaviors violating the requirements. Such requirements are usually composed of a safety part, whose run-time monitoring is well studied, and a liveness part, whose run-time monitoring is elusive. Most commonly, run-time monitors extract a safety property, that either over-approximates or under-approximate the original requirements, and thus tend to have a have high level of inaccuracy.A novel class of methods is proposed that employs randomization for monitoring the liveness properties. Roughly speaking, the method occasionally tosses a coin to determine whether to give up achieving a liveness requirement. Such methods are highly accurate and enjoy the property of graceful degradation: the longer a liveness property is not satisfied, the more likely is it to be rejected by the monitor. The project also investigates novel methods for evaluating the accuracy of monitoring approaches to compare various algorithms. Finally, monitors are usually ?passive? -- they only observe the computation, but do not participate in them ? but this work explores ?active? monitors that cooperate with the component to generate a computation that satisfies the requirements.