# Both ends of artificial intelligence impacting privacy: a review of violation and protection

## TL;DR

This paper reviews how AI can both threaten and protect privacy, analyzing 94 studies and using a graph database to visualize complex relationships.

## Contribution

A novel graph-based approach using Neo4J is introduced to model and visualize AI and privacy interactions.

## Key findings

- AI poses privacy threats through inference risks and data exploitation.
- AI can enhance privacy using techniques like federated learning and differential privacy.
- Interdisciplinary collaboration is needed to address regulatory and ethical challenges.

## Abstract

The intersection of Artificial Intelligence (AI) and privacy presents both significant challenges and opportunities. As AI systems become increasingly embedded in many aspects of our lives, including healthcare, finance, and social networks, and introduce significant concerns regarding privacy issues – the need for effective privacy-preserving mechanisms also grows. This review systematically analyzes 94 research papers in the field of AI and privacy. To model this complex issue, we categorized privacy in AI through a multi-dimensional approach that includes technological domains' privacy actions, privacy-preserving strategies, and AI-privacy interaction directions. A novel technique based on a Graph Database (Neo4J) which is available to the reader was employed to facilitate visualization of the complex relations between the reviewed objects. Moreover, the Graph, which is actually the review, can be queried and updated with future publications. Key findings indicate that AI can be both a potential threat to privacy, for example due to inference risks and data exploitation, as well as a tool for enhancing privacy through techniques such as federated learning and differential privacy. The study highlights regulatory, ethical, and technical challenges, emphasizing the need for interdisciplinary collaboration.