S-PoDL: A two-stage computational-efficient consensus mechanism for blockchain-enabled multi-access edge computing

Abstract

More recently, there has been a growing interest in the blockchain technique for emerging multi-access edge computing (MEC) applications in relation to security and privacy, considering the benefits of using it in edge computing. In some blockchain-enabled MEC using the best-known consensus algorithm proof-of-work (PoW), the computational efforts increase dramatically with the number of transactions, since a large amount of computational tasks should be conducted by miners with the limited computing resources in edge devices. Hence, improving the computational performance as an important but challenging issue in the design of blockchain system for MEC applications, has attracted intensive attention within last years. To further improve the performance of PoW algorithm, we present a novel implementation mechanism in this paper. Here, motivated by proof-of-deep-learning (PoDL) method in which the deep learning algorithm is used to maintain blockchain, through the design of a two-stage model to achieving computational tasks in PoDL-based blockchain systems, a novel computational-efficient consensus mechanism, named separate-proof-of-deep-learning (S-PoDL), is accordingly proposed. Thus, an energy-efficient blockchain-enabled MEC could be developed with our proposed S-PoDL, which arranges miners to carry out a two-stage-based computation on the basis of accounting_queue technique, while presenting the achieved models as proofs in MEC network. The comparative experiments are conducted between S-PoDL and PoDL, and the experimental results verify the feasibility and efficiency of our consensus mechanism S-PoDL for some blockchain-enabled MEC applications in relation to PoW-based cryptocurrencies, while effectively reducing computing burden of edge devices.

Introduction

While multi-access edge computing (MEC) as a new paradigm now addresses these ever-increasing demands from cloud computing service by processing data at the nearest available edge nodes and providing additional resources, i.e., storage space and computing power, at the edge side, then the low-latency requirement of computing services could be fulfilled, while effectively alleviating the loads of cloud server [1], [2]. However, there remain some technical challenges that should be addressed to fully exploit the advantages of MEC, during the design and implementation process of MEC [3]. Generally, when MECs are used in some typical industrial scenarios to provide a flexible and rapid deployment service at the edge nodes, the issue of data privacy and security should be handled, especially considering that the increasingly complex business environments make edge computing technology inapplicable in reliable and trustworthy applications. Then, the blockchain identified by researchers and practitioners as a promising technology, can deal with trust and security concerns, while providing a significant solution for emerging MEC applications [4], [5], [6], [7], [8], [9], where the edge devices are equipped with blockchain clients.

Currently, the effective combination of blockchain technology and MEC significantly improves the computational performance of MEC, considering that the requirements of interoperability and trust make the blockchain applicable to MEC platform [10], [11]. For example, as noted in [12], the integration of blockchain networks and MEC could be applied to some scenarios, e.g., cellular networks, then, a unified roaming-based blockchain framework was developed for cellular networks and its suitability was also demonstrated. Meanwhile, a blockchain technology was offered to present a trusted collaboration model between edge servers in an application scenario of MEC, where a permissioned blockchain framework was designed to provide a trusted service [13]. However, while applying blockchain to MEC, the resource-constrained transactions also impose very challenging obstacles to the deployment of blockchain-enabled MEC [12].

Generally speaking, blockchain was originally designed for the digital currency Bitcoin [14]. Then, cryptocurrency has been paid more and more attention. In the past decade, there are many new cryptocurrencies, including Burstcoin [15], Gridcoin [16], and FoldingCoin [17]. Among the available cryptocurrencies, most of them distribute rewards to miners according to their contribution. Meanwhile, in the technology architecture of cryptocurrencies, the consensus mechanism is used in blockchain with a decentralized way to reach agreement for transactions among individual users.

With the development of blockchain technology, various consensus algorithms have been proposed, such as proof-of-space (PoSpace), proof-of-activities, proof-of-work (PoW), proof-of- stake (PoS), and some others [14], [18], [19]. However, most of these consensus protocols may not be well employed in some practical environments due to too many competition and scalability limitations. Meanwhile, a large amount of electric power and computing resources including central processing unit (CPU), random access memory (RAM), and sensors, have been wasted to maintain blockchain system with these consensus protocols [20]. Moreover, the current blockchain framework using these consensus protocols for developing decentralized applications are greatly constrained by meaningless jobs. For example, Ethereum are hard to support varied decentralized application because of its redundantly ‘hashing’ procedure [21]. Among the available consensus algorithms, PoW has been used extensively. In PoW, it is hard to generate the piece of target data, but it is easy for other nodes to verify it. In recent years, some hybrid PoWs have become one of promising solutions for high performance blockchain. However, in PoW, each node in the network needs to calculate a hash value of the block header, which actually wastes a large amount of resources [22]. Therefore, more and more concerns for avoiding such limitation of wasting are arising gradually [23], [24], [25].

Currently, there are more and more consensus mechanisms being proposed to address the problem in PoW-based blockchain for MEC. For instance, PoS was designed to improve high energy consumption of PoW [26]. The PoS-based blockchain system can randomly select a node to generate the next block in accordance with the stake of nodes. However, when the PoS-based blockchain network scales out, the communication messages will grow explosively. The proof-of-capacity (PoC) was accordingly developed to save resources of PoW-based blockchians in Burstcoin [27]. PoC requires miners to dedicate an amount of disk space instead of computation in PoW. Although it saves computing resources, it wastes storage resources instead. In response to these limitations, the proof-of-deep-learning (PoDL) was proposed in which the deep learning algorithm was especially used to maintain blockchain [28], and PoDL can be deployed in any PoW-based blockchain MEC application for recycling energy. However, this method may also produce extra overhead, and when the proportion of dishonest nodes increases the overhead will rapidly grow. To recycle energy more effectively, we present a novel two-stage consensus mechanism, named separate-proof-of-deep-learning (S-PoDL). The S-PoDL divides all nodes in blockchain system into $N$ parts and accordingly establishes $N$ full_nodes. Consequently, the model_requester can release $N$ different deep learning models at the same time. Then, the nodes belong to different parts can train different models simultaneously. Additionally, those $N$ full_nodes can select the node with the highest accuracy from $N$ parts respectively. Finally, the $N$ nodes selected by full_nodes will be put into accounting_queue. Specifically, only the nodes in accounting_queue have accounting rights. Intuitively, our framework S-PoDL can effectively reduce the extra overhead and the computational time compared with PoDL.

Moreover, the issue of node cheat will lead to tampering with information in the blockchain network, thus destroying the property security and privacy in MEC. Then, to efficiently achieve security and privacy in MEC, two measures designed by PoDL in [28] have been also employed in our framework S-PoDL for prevention of node cheat. The first one is that the test_dataset are forbidden to be released until the second stage. This measure prevents nodes from training directly on the test_dataset to get higher scores in the first stage. The other measure is that the nodes must submit block header containing hash value of their trained models in the first stage for validation. It ensures that the model submitted by the node in the first stage is consistent with the one submitted in the second stage. Hence, through the use of above measures, the framework S-PoDL can efficiently achieve security and privacy in MEC as PoDL does.

Specifically, the contributions of this article are summarized as follows.

• We propose a more efficient consensus mechanism S-PoDL, through the use of multi-stage collaborative optimization strategy in which the miners are arranged to carry out a two-stage-based computation on the basis of accounting_queue technique, while presenting the achieved models as proofs in MEC network.

• The consensus mechanism S-PoDL can generate blocks faster and can be employed to blockchain-enabled MEC applications in relation to PoW-based cryptocurrencies with shorter block generation interval, comparing with traditional PoW.

• The consensus mechanism S-PoDL can significantly reduce extra overhead and more efficiently recycle energy compared with PoDL, moreover it can further save computing resources wasted by PoW.

The rest of this article is organized as follows. Section 2 provides a simple analysis of the backgrounds on PoW and PoDL. Section 3 details our consensus mechanism S-PoDL. Moreover, the computational performance of S-PoDL is theoretically analyzed in Section 4. Section 5 conducts the comparative experiments between the proposed model and other consensus mechanism, in an effort to verify the performance advantages of our method. Conclusion is provided in Section 6.