Executive Summary : | Data-driven artificial intelligence (AI) techniques are used in various applications such as healthcare, surveillance, behavior profiling, biometrics, legal practices, banking, and shopping. These AI techniques collect data from citizens to train machine learning or deep learning models that can capture behavioral patterns and traits. However, these models leave data footprints that can be extracted through various data inference methods. Data leaks can harm individuals and cause damage at societal and national levels. Data protection regulations like EU GDPR and CCPA have been developed to protect citizens' data from data mishandling. Companies must comply with "the right to be forgotten" and can delete their data at any point. Machine unlearning (MUL) algorithms are developed to efficiently erase information from trained ML/DL models. This project aims to address several key challenges, including the need for stronger theoretical foundations, addressing MUL challenges in different types of applications, addressing data availability, and developing standardization and benchmark datasets. The proposed solutions will work under different scenarios of data availability and provide confidence to governments and companies to comply with Data Regulation Policies to control digital data footprints on ML/DL systems. The developed MUL solutions can be applied in other areas such as security systems, banking, consumer data management, and medical data management. By developing effective MUL solutions, governments and companies can better manage their digital data footprints and comply with data regulation policies. |