University of Michigan – Dearborn

Calibrating Pedestrians’ Trust and Situation Awareness During Communication with Highly Automated Vehicles Under Dynamic Conditions Through the Design of an External Human-Machine Interface
Dania Ammar, Industrial and Systems Engineering

This dissertation contributes to understanding pedestrians’ safety issues and how technology solutions help. One challenging factor in introducing high-level automated vehicles (AVs) is their interaction with pedestrians. This study presents a user-centered approach for developing scalable external human-machine interfaces (e-HMIs) serving as communication means between AVs and pedestrians to improve their trust and situation awareness. In this work, a set of state-of-the-art methods are employed. A thorough literature review is first conducted to examine existing e-HMI solutions. Secondly, a user-centered design is incorporated to collect user needs related to different e-HMI attributes (e.g., modality, information type, and install locations). Consequently, e-HMI solutions are developed, and experiments are designed. Finally, testing is conducted in a virtual reality environment to examine the impact of developed e-HMIs on pedestrians’ trust and situation awareness levels. Data mining techniques and statistical modeling are applied, and results are interpreted accordingly.

Defenses Against Emerging Malware Threats
Nada Lachtar, Computer and Information Science

The popularity of cryptocurrencies has garnered interest from cybercriminals, spurring an onslaught of cryptojacking campaigns that aim to hijack computational resources for the purpose of mining cryptocurrencies. We present a cross-stack and application agnostic cryptojacking defense system. Our solution is resilient to multi-threaded and throttling evasion techniques that are commonly employed by cryptojacking malware. Finally, our implementation shows minimal performance impact while running a mix of benchmark applications. Furthermore, the unprecedented growth in mobile systems has transformed the way we approach everyday computing. Unfortunately, ransomware poses a great threat to consumers of this technology. We present RansomShield, an energy efficient solution that leverages CNNs to detect ransomware. We show that converting native instructions from Android apps into images using visualization techniques enable reliable detection. We evaluate the suitability of different models for mobile systems by comparing their energy demands using different platforms. Finally, we evaluate the robustness of this solution against adversarial machine learning attacks.