Publications
Abstract
Social media lets people with mental health issues share their experiences and find support online. But some need proper clinical help, making things worse. Checking what users post online can provide targeted care and reduce mistakes. Language models can analyze social media and classify mental health risks. We suggest a way to do this using a question-answering approach and protecting user data with privacy measures. Our results show this works well and could lead to better privacy-focused tools for mental health assessment.
Abstract
The paper explains how someone can understand a user's actions from their Wi-Fi activity without breaking encryption. It shows that an outside observer can figure out what's being sent even without direct access. When we use apps on our phones, they create patterns in network traffic. This pattern gives away some info, like when apps update. While encryption protects the data content, these patterns can still leak details like packet size and timing. These patterns can be used to steal private information, even if data is encrypted. These issues concern wireless communications, especially long-range ones, as outsiders can track and analyze this data.