Satwik ram Kodandaram, Mohan Sunkara, Sampath Jayarathna and Vikas Ashok
In the digital landscape, advertisements play a pivotal role, but their impact on blind users navigating websites with screen readers has been a largely unexplored territory. Through interviews with 18 blind participants, our study revealed a significant challenge: blind users are often misled by ads seamlessly integrated into the web page content. Conventional ad blockers face resistance, compelling us to devise an algorithm for automatic identification of contextually deceptive ads. Our multi-modal model, incorporating both handcrafted and automatically extracted features, demonstrated remarkable effectiveness with F1 scores of 0.86 and 0.88 on test datasets and real-world websites, respectively. This research sheds light on the nuanced experiences of blind users and introduces a practical solution to enhance their online interactions by addressing the deceptive nature of visually designed ads.











































