Design and Development of AI-Enabled Framework for IoT Networks

物联网（IoT）正在对我们的日常生活产生深远的影响，并将成为我们关键基础设施的基础，以及新兴和未来智能服务的基础。它是一个具有计算、通信、传感和驱动能力的设备网络，这些设备在反馈回路中相互作用，并有可能进行人为干预。随着物联网的继续扩散，拒绝服务等攻击也在继续，这些攻击可以让黑客访问流媒体视频源或控制自动驾驶汽车，因此物联网设备的预期行为可能会被证明是灾难性的。物联网应用领域的多样性导致了许多独特的挑战，我的研究计划将通过借鉴机器学习，安全和隐私，中间件和人机接口等学科来解决机器智能，安全和人在回路中的挑战，为物联网应用的安全开发和分析开发一个框架，以促进物联网网络的无处不在的部署。物联网网络涉及大量设备，这些设备以丰富而复杂的方式与物理世界、人类用户和云进行交互。这将带来几个挑战，包括：可扩展性，不断增加的资源需求协议，如设备上的视频处理，边缘设备的作用，开放的网络接口，以及所有设备之间的安全通信，同时保持用户及其数据的隐私。最近的新闻强调了网络汽车和植入式医疗设备的许多攻击机会，大众汽车被发现在排放测试中编程软件作弊的案例揭示了物联网设备歪曲自己的可能性。数据被实时收集和处理，关键决策基于这些数据（例如，例如，在自动驾驶汽车中打开灯或空调，或踩刹车），但数据可能由于以下几个原因而无效：传感器故障，恶意用户注入不正确的数据，或机器学习算法实施不正确。为此，本研究的目标是调查和开发解决上述挑战的解决方案，并将其纳入物联网网络开发和分析的框架，包括：机器学习算法作为可重复使用的组件，用于分析网络流量以进行异常活动检测和预测，应用层和网络层的新型安全机制，并评估机器智能准确性和性能、用户体验以及安全性和隐私之间的权衡。拟议框架的开发具有直接的经济和社会影响，这将保护物联网网络，从而保护企业和公民免受网络安全攻击和隐私侵犯风险。我的研究计划将在机器学习，中间件开发和物联网安全方面培训HQP，以满足这些领域对熟练劳动力的需求。

The Internet of Things (IoT) is having a profound impact on our daily lives, and will be the foundation of our critical infrastructure and the basis for emerging and future smart services. It is a network of devices with computing, communication, sensing and actuating capabilities that interact in a feedback loop with the possibility of human intervention. As IoT proliferation continues, so do attacks such as denial of service that provide hackers access to streaming video feeds or taking control of autonomous vehicles, and hence a breach in the intended behavior of IoT devices could prove catastrophic. The diversity of IoT application domains lead to many unique challenges, and my research program will address machine intelligence, security, and human-in-the-loop challenges by drawing from the disciplines of machine learning, security and privacy, middleware, and human-machine interfacing to develop a framework for the secure development and analysis of IoT applications to facilitate the ubiquitous deployment of IoT networks. IoT networks involve a large number of devices that interact with the physical world, human users, and the cloud in rich and complex ways. This will raise several challenges including: scalability, increasing resource-demanding protocols such as video processing on devices, the role of edge devices, open network interfaces, and the secure communication between all devices while maintaining the privacy of users and their data. Recent news highlighted many opportunities for attacks on networked cars and implanted medical devices, and the case where Volkswagen was found to have programmed their software to cheat on emission tests revealed the potential for IoT devices to misrepresent themselves. Data is collected and processed in real-time, and critical decisions are based on this data (e.g., turn on the lights or an air conditioner, or apply brakes in a self-driving car), but data can be invalid for several reasons including: sensor failure, malicious users injecting incorrect data, or machine learning algorithms implemented incorrectly. To this end, the objective of this research is to investigate and develop solutions that address the above challenges into a framework for the development and analysis of IoT networks, including: machine learning algorithms as reusable components for analysis of network traffic for anomalous activity detection and prediction, novel security mechanisms at the application and network layers, and evaluate the tradeoffs between machine intelligence accuracy and performance, user experience, and security & privacy. The development of the proposed framework has a direct economic and societal impact, which is securing IoT networks and thus protecting enterprises and citizens from cyber security attacks and privacy invasion risks. My research program will train HQP in machine learning, middleware development, and IoT security to meet the demands for a skilled workforce in those areas.