From Data to Defense: Designing Social Cyber Vulnerability Measures to Protect Older Adults Online
Researchers will develop a method to use the Social Cyber Vulnerability Index (SCVI) to detect fraud in online social spaces while factoring in demographic behavioral vulnerabilities that introduce representation bias due to disparities in detection-model training.
Funded by the CCI Hub
Project Investigators
- Lead Principal Investigator (PI): Hemant Purohit, George Mason University School of Computing
- Mason Co-PI: Fengxiu Zhang, Mason’s Schar School of Policy and Government
- Virginia Tech PI: Jin Hee Cho, Virginia Tech Computer Science Department
- Virginia Tech Co-PI: Chang-Tien Lu, Virginia Tech Computer Science Department
- Collaborator: Michin Hong, Indiana University School of Social Work
Rationale and Background
Assessing cyber vulnerability is pivotal in safeguarding online spaces against cyber threats and informing policy decision-making and intervention programs for cybersecurity. Programs such as the Common Vulnerability Scoring System (CVSS) can help assess technical vulnerabilities in information systems designed to protect users during online browsing. Yet, such methods do not account for the multifaceted behavioral vulnerabilities of all demographics. This gap highlights the need to design novel measurement tools to assess social cyber vulnerability.
Methodology
Researchers will:
- Re-annotate existing datasets for social media fraud, with a focus on identifying and labeling classes (romance scam, medicare scam) in posts aimed at specific demographic groups such as older adults.
- Examine stereotypical scam theme classes and contrast them with counter-stereotype scams (caregiver scams, child care scams), then develop an inclusive fraud-detection model.
- Construct an SCVI and develop a tool to identify risks and offer strategies to reduce vulnerabilities.
Projected Outcomes
By creating novel datasets for multimodal scam content on social media and developing inclusive fraud detection algorithms that account for demographically relevant scam types, this study will create an essential component for a comprehensive SCVI metric and tool.
This metric-based geo-mapping tool will enable the visualization of vulnerability across communities, allowing policymakers to prioritize resources and develop strategic programs to mitigate social cyber threats to vulnerable populations.