STARSS: Small: Defending Against Hardware Covert Timing Channels

保护存储在计算机系统中的敏感用户信息是一个迅速增长的问题，特别是当计算机被普遍使用在从国防到移动电话的各个地方时。恶意黑客已经找到了不择手段的方法来窃取敏感信息，主要是利用现有硬件和软件的漏洞。在多种形式的信息泄露中，隐蔽定时通道利用系统资源的访问定时，将具有较高安全证书的木马进程的机密信息泄露给具有较低安全证书的间谍进程。这种隐蔽的时序通道是难以捉摸的，因为它们通过时序调制间接通信，而不会在内存中留下任何物理痕迹。通过检测隐蔽时间通道的存在并对其进行防御，我们的项目通过防止不希望的敏感数据泄露给恶意方，为社会提供了显著的好处。该项目的三个中心目标是：1。研究硬件隐蔽时序通道的构造和运行背后的科学。设计检测硬件时序通道的机制；安装具有成本效益的运行时防御策略，以破坏隐蔽定时信道的可靠性。我们研究了识别关键指标事件的科学方法，并分析了它们在隐蔽时序通道运行中的作用。我们设计了捕捉运行时系统行为的机制，并检测隐蔽定时通道的存在。最后，为了防止信息泄露，我们使用分层、多层次的防御策略，通过测量防御来估计对手的敏捷性。部署后的有效性，并自动重新校准我们的防御，以摧毁定时通道的活动。

Safeguarding sensitive user information stored in computer systems is a fast growing concern, especially as computers are universally used everywhere from national defense to mobile phones. Malicious hackers have found unscrupulous ways to steal sensitive information largely by exploiting the vulnerabilities in existing hardware and software. Among the many forms of information leakage, covert timing channels exfiltrate secrets from a trojan process with higher security credentials to a spy process with lesser credentials by exploiting the access timing of system resources. Such covert timing channels are elusive since they communicate indirectly through timing modulation without leaving any physical trace in memory. By detecting the presence of covert timing channels and defending against them, our project offers significant benefits to society through preventing undesired sensitive data leakage to malicious parties. The three central objectives of this project are to: 1. Investigate the science behind the construction and operation of hardware covert timing channels, 2. Devise mechanisms to detect hardware timing channels, 3. Mount cost-effective, runtime defense strategies to undermine the reliability of covert timing channels. We investigate scientific ways to identify key indicator events, and analyze their roles in the operation of covert timing channels. We design mechanisms to capture the runtime system behavior, and detect the presence of a covert timing channel. Finally, to prevent information leakage, we use hierarchical, multi-level defense strategy where we estimate the agility of the adversary by measuring the defense?s effectiveness after deployment, and automatically recalibrate our defense to decimate the timing channel activity.