1: \begin{abstract}

2: Recommenders are central in many applications today. The most effective recommendation schemes, such as those based on \ac{CF}, exploit similarities between user profiles to make recommendations, but potentially expose private data.  

3: Federated learning and decentralized learning systems address this by letting the data stay on user's machines to preserve privacy: each user performs the training on local data and only the model parameters are shared. However, sharing the model parameters across the network may still yield privacy breaches. 

4: In this paper, we present \rex{}, the first enclave-based decentralized CF recommender.

5: \rex{} exploits \Ac{TEE}, such as Intel \ac{SGX}, that provide shielded environments within the processor to improve convergence while preserving privacy.

6: Firstly, \rex{} enables raw data sharing, which ultimately speeds up convergence and reduces the network load. 

7: Secondly, \rex{} fully preserves privacy. We analyze the impact of raw data sharing in both \ac{DNN}  and \ac{MF} recommenders and showcase the benefits of trusted environments %

8: in a full-fledged implementation of \rex{}.

9: Our experimental results demonstrate that through raw data sharing, \rex{} significantly decreases the training time by $18.3 \times$ and the network load by 2 orders of magnitude over standard decentralized approaches that share only parameters, while fully protecting privacy by leveraging trustworthy hardware enclaves with very little overhead.

10: \end{abstract}

11: