Adaptive Obfuscation Techniques for Privacy- Preserving Machine Learning in IoT Edge De- vices

Title	Adaptive Obfuscation Techniques for Privacy- Preserving Machine Learning in IoT Edge De- vices
Summary	This master thesis project focuses on developing an adaptive obfuscation frame- work for protecting multi-modal data in resource-constrained IoT environments.
Keywords
TimeFrame
References
Prerequisites
Author	Mahdi Fazeli
Supervisor	Mahdi Fazeli
Level	Master
Status	Open

This master thesis project focuses on developing an adaptive obfuscation frame- work for protecting multi-modal data in resource-constrained IoT environments. With the growing reliance on IoT systems in critical domains such as health- care, autonomous vehicles, and industrial automation, ensuring data privacy is a major challenge. Current obfuscation methods, like ObfNet, are limited in their flexibility and vulnerable to adversarial attacks that can expose sensitive information. The proposed project will address these issues by designing a framework that dynamically adjusts obfuscation levels based on data complexity and risk factors, enhancing its resilience against adversarial attacks. Additionally, it will create a privacy-measuring metric to evaluate the effectiveness of obfuscation techniques, balancing privacy, security, and performance in real-world applica- tions. The project will aim to optimize obfuscation techniques for IoT devices with constrained computational and energy resources, making it feasible for deployment in sectors such as healthcare and autonomous systems.