Support for security and safety of programmable IoT systems

This project is a collaborative grant involving the University of Michigan (Award number CNS-1646392) and the University of Illinois at Urbana-Champaign (Award number CNS-1646305). The major goals of the project are to improve safety and security in Internet of Things (IoT) systems where multiple devices (things), each designed in isolation from others, are brought together to form a networked system, controlled via one or more software applications

("apps"). The project will help develop techniques for testing IoT apps efficiently and for enforcing safety and security constraints on their run-time behavior. Moreover, we are experiencing new kinds of interactions between the cyber and physical world. For example, autonomous driving vehicles need to observe the actual physical world, e.g., using computer vision to understand traffic signs. Some contributions of the work include an open source system, FlowFence, that supports fine-grain information flow policies on Android; using risk-based grouping of user permissions rather than functional grouping to improve security risk/usability tradeoff in IoT systems; and open source adversarial testing software for generating robust physical perturbation attacks on computer vision classifiers and detectors. More information at https://iotsecurity.engin.umich.edu.