Multi-Party Computation in IoT for Privacy-Preservation
Abstract: Preservation of privacy has been a serious concern with the increasing use of IoT-assisted smart systems and their ubiquitous smart sensors. To solve the issue, the smart systems are being trained to depend more on aggregated data instead of directly using raw data. However, most of the existing strategies for privacy-preserving data aggregation, either depend on computation-intensive Homomorphic Encryption based operations or communication-intensive collaborative mechanisms. Unfortunately, none of the approaches are directly suitable for a resource-constrained IoT system. In this work, we leverage the concurrent-transmission-based communication technology to efficiently realize a Multi-Party Computation (MPC) based strategy, the well-known Shamir's Secret Sharing (SSS), and optimize the same to make it suitable for real-world IoT systems.
Paper Reviews
- What are major strengths?
- The studied problem is important.
- What are shortcomings?
- The key novelty of this paper could be better elaborated.
- What are major strengths?
- It is relevant to this conference.
- The presentation is good.
- The performance of the proposed scheme was evaluated by implementations in Cooja as well as IoT/WSN testbeds DCube and FlockLab composed of 45 and 24 TelosB motes, where the algorithms were implemented in Contiki OS for TelosB motes.
- The paper shows that ReLI can operate up to 80% faster and consume up to 78% less radio-on time compared to the traditional implementation of the strategy in a publicly available IoT/WSN testbed containing 45 nodes.
- What are shortcomings?
- In abstract, too many backgrounds are described.
- In introduction, too many backgrounds are described. To the work of the author, it does not declare clearly.
- In introduction, too much introductions are about SSS and MiniCast, but to how to integrate them are not clear.
- There is not significant contribution of this work. It is just an attempt to integrate SSS and MiniCast, and the results show that it is just can afford few nodes. It is not practical.