Evaluating Apple Intelligence'S Writing Tools for Privacy against Large Language Model-Based Inference Attacks: Insights from Early Datasets

The misuse of Large Language Models LLMs to infer emotions from text for malicious purposes, known as emotion inference attacks, poses a significant threat to user privacy. In this paper, we investigate the potential of Apple Intelligence's writing tools, integrated across iPhone, iPad, and...