Design Challenges for Scam Prevention Tools to Protect Neurodiverse and Older Adult Populations

Scams through various communication mediums whether online social media, emails, search ads, or offline via phone calls and short messages (SMS) have dramatically increased after the pandemic. Prior technical research has leveraged artificial intelligence (AI) algorithms for scam detection tasks, such as spam and phishing detection, to design systems for preventing users from scams. However, due to the heavy data- dependence on AI techniques, in particular machine learning, the training datasets for classification models could be non-representative in capturing different nuances and evolution of social engineering attacks that aim to target vulnerable populations like the elderly or neuro-diverse individuals. For such populations, biases in the resulting models, in turn, lead to vulnerabilities in associated cybersecurity tools. Further, a one-size-fits-all approach to create interaction mechanisms in such tools reduces their value to protect individuals with neurodiverse profiles and older adults with varying cognitive abilities. In this paper, we synthesize the existing literature with the goal of identifying the sources of inclusive design challenges for scam filtering and prevention tools through a critical analysis of literature on both current solutions explored in cybersecurity research and the needs of individuals with diverse disability profiles. We present an Inclusive AI-driven Cybersecurity (IAC) Framework for designing effective and accessible tools to protect all populations. The findings of this research can inform effective designs of scam prevention tools across different communication media with the inclusive goal of ultimately protecting all populations.