A proof of contribution in blockchain using game theoretical deep learning model

Jin Wang

2024-08-25

Abstract:Building elastic and scalable edge resources is an inevitable prerequisite for providing platform-based smart city services. Smart city services are delivered through edge computing to provide low-latency applications. However, edge computing has always faced the challenge of limited resources. A single edge device cannot undertake the various intelligent computations in a smart city, and the large-scale deployment of edge devices from different service providers to build an edge resource platform has become a necessity. Selecting computing power from different service providers is a game-theoretic problem. To incentivize service providers to actively contribute their valuable resources and provide low-latency collaborative computing power, we introduce a game-theoretic deep learning model to reach a consensus among service providers on task scheduling and resource provisioning. Traditional centralized resource management approaches are inefficient and lack credibility, while the introduction of blockchain technology can enable decentralized resource trading and scheduling. We propose a contribution-based proof mechanism to provide the low-latency service of edge computing. The deep learning model consists of dual encoders and a single decoder, where the GNN (Graph Neural Network) encoder processes structured decision action data, and the RNN (Recurrent Neural Network) encoder handles time-series task scheduling data. Extensive experiments have demonstrated that our model reduces latency by 584% compared to the state-of-the-art.

Cryptography and Security,Machine Learning

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

This paper aims to address several key issues in edge computing resource management:

1. **Resource Limitation**: A single edge device cannot handle the various complex computing tasks in a smart city, thus requiring large-scale deployment of edge devices from multiple service providers to build a resource platform.
2. **Resource Scheduling and Allocation**: Choosing the computing power of different service providers is a game theory problem. How to incentivize service providers to actively contribute their valuable resources and provide low-latency collaborative computing capabilities is the core challenge of this paper.
3. **Limitations of Traditional Methods**: Traditional centralized resource management methods are inefficient and lack credibility. Introducing blockchain technology can achieve decentralized resource trading and scheduling, significantly improving transaction transparency, efficiency, and resource utilization.
4. **Improvement of Consensus Mechanisms**: Existing consensus mechanisms (such as Proof of Work (PoW) and Proof of Stake (PoS)) have certain limitations and cannot fully incentivize participants' contributions to the community. Therefore, this paper proposes a Proof of Contribution (PoC) mechanism to maximize contributions to the entire edge computing community.

### Solutions

To address the above issues, the paper proposes the following solutions:

1. **Game Theory-Based Deep Learning Model**: By introducing game theory and deep learning models, consensus among service providers in task scheduling and resource allocation is achieved. The model consists of a dual encoder (Graph Neural Network (GNN) encoder and Recurrent Neural Network (RNN) encoder) and a single decoder, where the GNN encoder processes structured decision action data, and the RNN encoder processes time-series task scheduling data.
2. **PoC Consensus Mechanism**: A Proof of Contribution (PoC) mechanism is proposed to select resource inputs and task resource providers participating in computing and competition to maximize contributions to the community. The PoC mechanism determines the influence of participants in the network based on their computing power and resource levels, thereby improving the performance efficiency of the edge network.
3. **Knowledge Distillation**: To address the issue of limited resources on edge devices, a knowledge distillation method is used to compress complex teacher models into simpler student models. The student models perform on-chain PoC transactions through peer-to-peer communication and decentralized consensus on distributed edge nodes.

### Main Contributions

1. **Proposing the PoC Consensus Mechanism**: Selecting resource inputs and task resource providers participating in computing and competition to maximize contributions to the community while minimizing average task delay. All community resource trading records are recorded on the blockchain to ensure transaction transparency and fairness.
2. **Dual Encoder Single Decoder Model**: Combining the dual encoder structure of GNN and RNN, it can comprehensively extract task information and time-series information to generate actual resource allocation plans or game decisions. This model can consider both global and local features simultaneously, better modeling the strategy generation process in the game.
3. **Diverse Game Behaviors**: Simulating diverse game behaviors through deep learning networks, ensuring the system can still operate normally even if some nodes fail or behave maliciously. Diverse game behaviors align with the core idea of Byzantine fault tolerance, ensuring the secure and stable operation of the blockchain system.
4. **Application of Knowledge Distillation**: Using knowledge distillation methods to compress complex teacher models into simpler student models, which perform on-chain PoC transactions through peer-to-peer communication and decentralized consensus on distributed edge nodes, addressing trust and privacy issues in centralized management.

### Experimental Results

Experimental results show that compared to existing state-of-the-art methods, this model can reduce latency by 5.84%, verifying its effectiveness and superiority in resource management and task scheduling.