Resilient, Secure, and Programmable Next-Generation Internet of Things (IoT)

在我们的家里、办公室和汽车里，我们周围都是嵌入式计算设备。随着世界朝着扩大连接的方向发展，这些设备连接到互联网的情况越来越普遍。这种互联互通为提高我们的生活质量提供了巨大的机会。例如，许多家庭已经拥有智能电表和智能恒温器等设备，这些设备可以从互联网上获取能源价格和天气预报的实时信息，以确保温度设置自动设置到最佳值。这些互联的嵌入式设备统称为物联网(IoT)。 不幸的是，物联网设备的互联特性使其容易受到意外故障和恶意攻击。因为我们的日常活动依赖于这些设备，它们的故障可能会削弱我们的日常生活，在某些情况下，还会导致毁灭性的后果。一个例子是Nest恒温器在例行软件升级后出现故障。更重要的是，这些设备是恶意攻击者的目标，他们想要为了经济利益或个人目的对社会造成伤害。最近，针对智能电表、汽车甚至心脏起搏器和胰岛素泵等医疗设备的安全攻击事件层出不穷。因此，迫切需要使物联网设备对意外故障和恶意攻击具有弹性。此外，由于物联网设备的激增，缺乏经验丰富的软件开发人员以可靠和安全的方式对这些设备进行编程。因此，我们需要提供可编程的抽象来支持这些设备的软件开发，而不需要高水平的开发人员的专业知识和努力。 在这份提案中，我将开发新的技术来增强物联网软件的可靠性、安全性和可编程性。这一问题之所以具有挑战性，有三个原因：首先，物联网设备和架构种类繁多，每种都有不同的限制和要求。设计一个可以跨多个设备工作的统一解决方案是具有挑战性的。例如，智能电表的约束与起搏器的约束非常不同。第二，不同于传统的计算机系统，物联网系统部署在没有人参与的情况下。因此，这些系统不能使用人工决策来解决模棱两可或错误警报。例如，当一辆车在高速公路上行驶时，错误地向司机发出可能的攻击警报可能会导致他/她恐慌。最后，由于物联网设备通常是廉价的大规模生产设备，设备制造商不愿采用大规模部署成本高昂的技术，因此成本非常重要。这项提案将采用四步走的方法来应对上述三项挑战。它还将在5年内培训15名HQP(10名本科生、3名博士、2名硕士)。

We are surrounded by embedded computing devices, in our homes, offices and cars. As the world moves towards expanding connectivity, it is increasingly common for these devices to be connected to the Internet. This connectivity opens up tremendous opportunities to improve our quality of lives. For example, many homes already have devices such as smart meters and smart thermostats, which can obtain real-time information about both energy prices and weather forecasts from the Internet to ensure that the temperature settings are automatically set to their optimal values. Collectively, these interconnected embedded devices are known as the Internet of Things (IoT). Unfortunately, the connected nature of IoT devices renders them vulnerable to both accidental faults and malicious attacks. Because we depend on these devices for our day-to-day activities, their failures can cripple our daily lives and, in some cases, lead to devastating consequences. An example is the failure of the Nest thermostat after a routine software upgrade. Even more importantly, these devices are targets for malicious attackers who want to wreak harm on society for financial gains or for personal ends. Recently, there have been numerous instances of security attacks on smart meters, automotives, and even on medical devices such as pacemakers and insulin pumps. Therefore, there is a compelling need to make IoT devices resilient to both accidental faults and malicious attacks. Further, because of the proliferation of IoT devices, there is a shortage of experienced software developers to program these devices in a reliable and secure fashion. Therefore, we need to provide programmable abstractions to enable software development for these devices, without requiring high-levels of developer expertise and effort. In this proposal, I will develop novel techniques to enhance the reliability, security and programmability of IoT software. This problem is challenging for three reasons: First, there are a wide variety of IoT devices and architectures, each with different constraints and requirements. Designing a unified solution that can work across multiple devices is challenging. For example, the constraints of a smart meter are very different from those of a pacemaker. Second, unlike traditional computer systems, IoT systems are deployed in situations where there is no “human in the loop”. Hence these systems cannot use human decision making to resolve ambiguities or false alarms. For example, when a car is driving down the highway, falsely alerting the driver of a possible attack may cause him/her to panic. Finally, because IoT devices are often inexpensive, mass-produced devices, device manufacturers are reluctant to adopt techniques that are costly to deploy at large scale, and hence cost is very important. This proposal will tackle the above 3 challenges using a four-step approach. It will also train a total of 15 HQPs (10 Undergrads, 3 PhDs, 2 Masters) over 5 years.