| * | # |
|---|---|
| Link | link. |
| topic | Integrating Blockchain With Artificial Intelligence for Privacy-Preserving Recommender Systems |
| Key words | implicit voting, collaborative filtering, blockchain, |
core point
1. The main problems and solutions to be solved in the article:
1) Problems to be solved
Recommendations are generated in two ways: content-based filtering and collaborative filtering. In collaborative filtering, a list of predictions is generated by determining the correlation between user history and other user interests [6]. On the other hand, item and user profile descriptions are explored in content-based filtering. Here, user profiles are constructed based on the user's history and user's ratings [7].
In order to make the best recommendations using collaborative filtering [7], [8] companies store personal data of their customers. Therefore, user private data leakage incidents are prone to occur [9].
2) The main work of the article
We design a framework named Private-Rec, a user-centric recommender system utilizing collaborative filtering. The entire process of data collection and storage is done by our platform without sharing data with companies. In our proposed Private-Rec, no entity is able to access user data, and computations are suggested to be performed in a secure manner. Hence, these companies will not have any chance of accessing user data. Whenever user data is used by our platform, the transaction records of the data sharing are stored on the blockchain. Users of our platform will be rewarded by the platform when using their data. Blockchain ensures that no user data is used to calculate recommendations without incentives.
In this paper, our contributions are as follows:
we proposePrivate-Rec, an AI-based privacy-preserving recommendation system that ensures user data privacy.
we useblockchainto store data transactions, holding companies accountable.
Private-Rec Guaranteeaccountability, integrity, pseudonymity and privacy. Data privacy concerns have been addressed by storing all data in a responsible data cloud, our platform usesencryptionfunction to ensure privacy.
introducedCluster-Based Incentive Mechanism. Users who will share data to generate recommendations will be rewarded with some points. These points can later be used in the platform.
Three algorithms are introduced for processing: request sending mechanism for guest users, recommendation generation mechanism, join request management and incentive mechanism. The platform has been evaluated in different recommender systems.
2. Content of the article
Voting Techniques
Explicit Voting - User ratings of preferences for specific items.
Implicit voting - will record user actions in the system to generate recommendations (recording user data has privacy issues)
Algorithm:
memory-based algorithm√