PUBLISH-SUBSCRIBE MODEL ARCHITECTURE FOR DECENTRALIZED INTERNET OF THINGS USING TANGLE

• Main Author: Septianus Darmawan, Felix
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/48555
• Institution: Institut Teknologi Bandung

Centralized architecture has weaknesses in the form of a single point of failure and the risk of data tampering by malicious admin. There exists a decentralized solution that solves these problems, which is blockchain, but blockchain has limited transaction throughput and confirmation rates. To meet the needs of IoT, a system that is better than blockchain is needed. From the existing advancement of blockchain, one technology was chosen in which the solution will be built. Tangle is the solution chosen because it has better scalability than blockchain. Tangle uses the directed acyclic graph structure to improve the performance of blockchain. In this final project, a publish-subscribe model is built on top of Tangle. The system is made secure in terms of confidentiality, authenticity, integrity, and availability by using encryption, digital signatures, and the decentralized nature of the blockchain. The system is made for the use case of purchasing data between organizations. The solution is built to run on a gateway. The use of the MQTT protocol for communication of IoT devices to the gateway makes this system usable by many resource-limited devices. In addition, by placing the system at a gateway, the sending of data to IoT devices that use the same gateway is not constrained by Tangle's performance. System performance is tested by measuring transaction throughput and delay between the generation of data and the reading of data. The solution built successfully sends data with a publish-subscribe system safely with better performance than conventional blockchain. However, the Tangle network is still currently centralized. Additionally, the delay has no upper limit value, so it cannot be used on an IoT system that is sensitive to timings. The maximum limit on Tangle's performance has not been tested due to limited resources.