AP4L Project: Transition Guardians

By Dr Shalini, University of Surrey

Major life events and transitions significantly shape a person's life, often requiring them to rebuild or redefine aspects of their identity. In times of identity change, individuals may turn to social media for support and guidance. When these transitions involve life-changing experiences, such as relationship breakups or coming out as LGBTQ+, people frequently share their stories on platforms like Reddit to connect with others and navigate the challenges they face. While such self-disclosures can be rewarding social interactions in a community of shared interests, they pose privacy risks and the threat of online harm. Social media facilitates sharing personal information that can inadvertently expose users to harm through piecing together contextual clues. Our research investigates techniques for detecting subtle privacy leaks in online narratives. 

We provide the following major contributions during our research:


  1. Research into the identification and retrieval of such risky self-disclosures of personal information identifiers (PIIs) is hampered by the lack of open-source labeled datasets, due to the high annotation costs and privacy risks associated with the curation of datasets containing self-disclosive text, especially by vulnerable populations, from platforms such as Reddit. To foster reproducible research into PII-revealing text detection, we develop a taxonomy of PII-revealing categories for vulnerable populations and introduce a synthetic PII-labeled multi-text span dataset generated from 3 text generation Large Language Models, prompted to resemble the original Reddit posts.

  2. The project is developing a browser extension to help users reflect on their digital traces and manage their online presence. This tool can highlight potential privacy risks by showing how small pieces of personal information, when combined, can lead to unintended exposure. Users can customize the extension to track PII leaks on specific platforms, such as Facebook, while excluding others like LinkedIn, giving them greater control over their online privacy.

  3. Further, we aim to enhance the functionality of this extension by providing users with statistical insights into the PII(s) they are revealing through their online interactions. For example, the plugin could inform users if, say, 10% of people in location X identify as transgender, highlighting how sharing this information may associate them with a minority or vulnerable group, potentially making them more identifiable. However, to ensure privacy, these aggregate statistics will be calculated using privacy-preserving techniques. This approach allows for useful insights without compromising the privacy of individuals included in the aggregated data.

Comments

Post a Comment

Popular posts from this blog

Welcome to the first of our monthly AP4L bulletins! AP4L has reached the halfway mark now in the three year process, and with some exciting developments already under our belts, and many more to come, we wanted to share our progress with you.  Each month we will be sharing with you some of our outputs, developments and latest news from a different area of the project. This month, I am going to give you an overview of AP4L, and next month, we will share more on our Public Involvement in Privacy (PIP) Panel, which is proving to be a pivotal part of AP4L.  So, what is the AP4L project all about? Well, sometimes, life “just happens”. Be it divorce, being diagnosed with a serious illness, leaving the armed forces, or deciding to come out as gay or transitioning gender, these moments in time can be very emotional, distressing and people do not typically have the “bandwidth” to stop and consider their online safety and privacy during these difficult life transitions.  In the case of a relatio
Read more
Image
  Public Involvement in Privacy (PIP) panel – what is it and how does it work ?   As highlighted in last month’s blog, the PIP panel is based on a model of public involvement that we have adapted from similar gold-standard practices in health and social care research (UK Standards for Public Involvement, National Institute for Health and Care Research (NIHR), 2019): https://www.invo.org.uk/wp-content/uploads/2019/11/UK-standards-for-public-involvement-v6.pdf   The standards highlight what high-quality public involvement should look like and ensures good practice when involving members of the public in research and a range of other activities. At their very core, the standards recommend involving people with relevant lived experience, and other research stakeholders as partners alongside the academic research team. The PIP panel exists within the AP4L project to ensure that those with lived experience or relevant life transitions are involved from the outset and are meaningfully
Read more

Engaging with hard-to-reach and potentially vulnerable participants in the AP4L Project

By Dr. Ryan Gibson and Prof. Wendy Moncur Introduction AP4L’s overarching goal is to develop privacy-enhancing technologies (PETs) that support people undergoing a significant life transition to be safer online. The first step in building PETs that make a real difference is to understand what these people do online and the subsequent challenges they face. Central to this process is the involvement of participants from the four populations we are exploring in AP4L - Leaving the Armed Forces; LGBTQIA+; Living with Cancer; Relationship Breakdown – who can share their lived experience of transitioning online.  These experiences may then be translated into design requirements for the development team to generate PETs that more accurately reflect the transition activities being conducted online and the related privacy risks and harms that people encounter.  Nevertheless, designing and recruiting for research involving people who have undergone a life transition is extremely difficult. For
Read more