When AI Meets Information Privacy: The Adversarial Role of AI in Data Sharing Scenario

Majeed, Abdul; Hwang, Seong Oun

doi:10.1109/access.2023.3297646

Cited by 9 publications

(1 citation statement)

References 58 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The use of synthetic data (SD) as a background knowledge (BK) element is still largely unexplored, despite these advancements. By putting out a fresh taxonomy that incorporates AI-generated data into the privacy discourse, our work aims to close this gap [6]. By means of this integration, the research illuminates the subtle dangers that SD poses in data publishing situations, enhancing our comprehension of privacy hazards in the age of artificial intelligence-driven data synthesis.…”

Section: Related Work a Threats Of Artificial Intelligence In Privacy...mentioning

confidence: 99%

AI & The Privacy Paradox: Navigating Synthetic Data Vaults

Manoj,

Binu

et al. 2024

IJRASET

View full text Add to dashboard Cite

Exploring the synergies and threats of artificial intelligence in the information privacy domain, the focus is on synthetic data (SD) as a form of background knowledge (BK) that can compromise privacy in data publishing scenarios. A novel taxonomy of BK that includes AI-generated data is proposed, demonstrating how it can lead to various kinds of privacy breaches, such as re-identification, sensitive attribute disclosure, and group privacy disclosure. Experiments are conducted on a real-life benchmark dataset using a state-of-the-art AI model to generate SD, showing that SD can pose serious risks to individual privacy when carefully crafted and used by adversaries. In realtime processing paradigms this method may face some challenges such as the highly customized implementation of libraries needed for computing statistics and visualizations of query answers for robust analysis without jeopardizing user’s privacy. To overcome the challenges in real-time processing paradigms, the algorithm can use stream anonymization methods or online learning techniques to anonymize data in an incremental and adaptive manner

show abstract

Section: Related Work a Threats Of Artificial Intelligence In Privacy...mentioning

confidence: 99%