Human activity profile tracking using static analysis of binary code access patterns for freeze-safe iot systems

IoT-driven systems perform human-interactive services based on communications that is constructed with irregular links between heterogeneous things. These links are major causes of propagating unwanted errors or states in entire systems. This paper focuses on reducing the number of possible cases of unwanted system halts by quickly propagating abnormal freezing statuses. Our approach is based on the early detection and mitigation of the uncontrollable error injections that cause these potential errors. Because human activity is always tightly coupled with internal software, the binary access patterns in software execution, which is triggered by human activity, can be abstracted with user-defined profile data. The expected profile, which is recorded in the pre-simulation step, will be compared with the runtime monitoring results of the state transitions. The proposed method is applied to the next-generation prototype of commercial devices, showing capability as case study toward freeze-safe IoT applications.