Personalized recommendations combat internet information overload. Privacy concerns hinder this; users hesitate to share data. We suggest generating fake profiles to protect privacy without compromising recommendation accuracy. Our client-based framework addresses this, ensuring existing algorithms remain unchanged. Our privacy model defines ideal fake profiles, emphasizing feature distribution similarity and sensitive subject coverage. We implement this using a product classification subject repository. Theoretical and experimental evidence supports our approach’s effectiveness.

• Fake Profiles for Privacy
• Personalized Recommendations
• Privacy Protection