[Paper Review] Trustworthy Edge Computing through Blockchains

Abdullah Al-Mamun, Dongfang Zhao|arXiv (Cornell University)|May 15, 2020

Blockchain Technology Applications and Security26 references4 citations

TL;DR

This paper proposes DEAN, a blockchain-based consensus protocol for edge computing that enables trustworthy data sharing under resource constraints by decentralizing trust across edge nodes. Leveraging parallel block processing and smart load balancing, DEAN achieves up to 25× higher throughput and 18× lower latency than Ethereum, Parity, and Hyperledger Fabric on 1,000-node networks, while maintaining high resilience with over 50% trusty nodes even under arbitrary failures.

ABSTRACT

Edge computing draws a lot of recent research interests because of the performance improvement by offloading many workloads from the remote data center to nearby edge nodes. Nonetheless, one open challenge of this emerging paradigm lies in the potential security issues on edge nodes and end devices, e.g., sensors and controllers. This paper proposes a cooperative protocol, namely DEAN, across edge nodes to prevent data manipulation, and to allow fair data sharing with quick recovery under resource constraints of limited storage, computing, and network capacity. Specifically, DEAN leverages a parallel mechanism equipped with three independent core components, effectively achieving low resource consumption while allowing secured parallel block processing on edge nodes. We have implemented a system prototype based on DEAN and experimentally verified its effectiveness with a comparison with three popular blockchain implementations: Ethereum, Parity, and Hyperledger Fabric. Experimental results show that the system prototype exhibits high resilience to arbitrary failures: the percentile of trusty nodes is much higher than the required 50\% in most cases. Performance-wise, DEAN-based blockchain implementation outperforms the state-of-the-art blockchain systems with up to $25 imes$ higher throughput and $18 imes$ lower latency on 1,000 nodes.

Motivation & Objective

• Address the lack of secure, scalable, and efficient consensus protocols tailored for resource-constrained edge nodes and end devices.
• Overcome the limitations of traditional blockchain systems—such as high storage, compute, and network demands—that are incompatible with edge environments.
• Design a decentralized, fault-tolerant protocol that ensures data integrity and resilience against arbitrary node failures in heterogeneous, wireless edge networks.
• Achieve high performance and low overhead through optimized load balancing across computation, storage, and network resources in edge deployments.
• Demonstrate that blockchain-based trust can be effectively adapted to edge computing by validating the protocol in a real system prototype with comparative benchmarks.

Proposed method

• Propose DEAN, a decentralized, edge-optimized blockchain protocol that partitions edge nodes into sensor and storage roles to balance load and reduce resource pressure.
• Implement a parallel block processing mechanism with three independent core components: block generation, verification, and storage, enabling concurrent execution and reducing latency.
• Use a hybrid consensus model inspired by PBFT and full replication, ensuring Byzantine fault tolerance with fewer messages than traditional PBFT while minimizing storage overhead.
• Apply data compression and shallow replication techniques to reduce storage requirements, making blockchain feasible on devices with limited memory and storage.
• Distribute transaction processing across sensor nodes and storage nodes to balance computational and network loads, improving scalability and fault tolerance.
• Design a dynamic load-balancing strategy that adapts to node heterogeneity and network conditions, smoothing performance across varying sensor-to-storage node ratios.

Experimental results

Research questions

• RQ1 Can a blockchain-based consensus protocol be effectively adapted to resource-constrained edge nodes without relying on high-end hardware or centralized trust?
• RQ2 How does the proposed DEAN protocol maintain high data fidelity and resilience under arbitrary node failures in a decentralized edge network?
• RQ3 To what extent can DEAN outperform existing blockchain systems like Ethereum, Parity, and Hyperledger Fabric in terms of throughput and latency on large-scale edge deployments?
• RQ4 How do varying ratios of sensor to storage nodes affect the performance and reliability of the DEAN protocol in real-world edge environments?
• RQ5 Can the combination of parallel processing and smart load balancing significantly reduce resource consumption while maintaining strong security guarantees in edge computing?

Key findings

• The DEAN-based blockchain prototype achieves up to 25× higher throughput and 18× lower latency compared to Ethereum, Parity, and Hyperledger Fabric on a 1,000-node edge network.
• The system maintains a trusty node ratio exceeding 50% in most configurations, demonstrating strong resilience against arbitrary node failures.
• Throughput and latency remain stable across different sensor-to-storage node ratios, with no significant performance degradation observed even at extreme ratios like 1:5.
• Larger sensor-to-storage ratios (e.g., 1:5) produce smoother performance curves due to reduced network heterogeneity, as sensor-to-storage links typically offer higher bandwidth than internal storage node networks.
• The performance trends are consistent and predictable across all tested node scales, indicating that DEAN scales effectively with increasing network size.
• The system prototype successfully handles high transaction loads with minimal variance, confirming the robustness of the parallel processing and load-balancing design.