The rapid development of IoT (Internet of Things) systems and cloud techniques has paved the way for recommender systems to facilitate the daily life of users. However, the accompanying cybersecurity risks, such as environmental attacks and software attacks, must not be ignored. Thus, the security problem in recommender systems becomes a serious challenge for cloud-based IoT services. Moreover, most of existing collaborative recommendation algorithms mainly focus on user-item interaction relationships but seldom consider user-user or item-item co-occurrence relationships, which may affect prediction accuracy. To overcome the above shortcomings, this paper proposes a security-driven hybrid collaborative recommendation method to deal with the large-scale IoT services accessible by clouds in a more scalable and secure manner. Our proposal integrates the factorization-based latent factor model with the neighbor-based collaborative model to mine not only user-service interaction relationships but also user-user and service-service co-occurrence relationships. Moreover, the local sensitive hash (LSH) technique is adopted to speed up the neighbor searching and preserve users’ sensitive information for security concerns based on hash mapping. Finally, experiment results demonstrate that the proposed method can improve prediction accuracy while guaranteeing information security.
|Journal||Computers and Security|
|Publication status||Published - Oct 2020|
Scopus Subject Areas
- Computer Science(all)
- Collaborative recommendation
- IoT services