Freshness and Security-Aware Cache Update in Blockchain-Based Vehicular Edge Networks

Vehicular Edge Computing (VEC) integrated with blockchain technology holds great promise for delivering secure temporal data services. However, ensuring data freshness in edge nodes remains a difficult task due to the temporal nature, heterogeneity, and privacy concerns associated with the data. Secondly, the dynamic VEC environment degrades blockchain performance, resulting in low transaction throughput or excessive energy consumption. As a result, we formulate the problem of Blockchain-based Edge Cache Update (BECU), which aims at maximizing both edge cache benefit and blockchain performance by optimizing cache decisions and critical blockchain parameters, including primary node, block size, and block interval. Furthermore, we develop the Contextual Multi-Armed Bandit for Caching Update (CMAB-CU) algorithm for online cache decision-making, which evaluates rewards by training a linear function based on mobility features and temporal data characteristics. Additionally, we design the Deep Q-learning Network for Blockchain Parameter Optimization (DQN-BPO), which dynamically determines blockchain parameters to strike the balance between transaction throughput and energy consumption. Finally, we conduct simulations using realistic vehicular traces, demonstrating that the proposed algorithms outperform the UCB and FBI algorithms in terms of edge cache benefit and blockchain performance by 95.56% and 144.93%, respectively.

I. Introduction

Intelligent vehicles are expected to be one of the most dominant areas of modern consumer electronics, thanks to recent advancements in data sensing, wireless communication and artificial intelligence technology. These intelligent vehicles are equipped with various sensors and communication devices [1], enhancing traffic safety and efficiency through information fusion from multiple sensors. Vehicular edge computing (VEC) is promised to provide temporal data service to vehicles at network edge. VEC supports various types of Intelligent Transportation Systems (ITSs), including automatic driving [2], vehicle trajectory prediction [3], and vehicle positioning [4]. However, the private information, e.g., the identity and location of vehicles and caching information of edge node, may be hijacked and tampered with by attackers during edge cache service [5]. To guarantee data security, blockchain technology is introduced to store cache transactions as a chain of blocks, which are irreversible once recorded and cannot be altered retroactively. Nonetheless, investigating an effective blockchain-based VEC system is challenging due to the following issues. 1)

Dynamically Edge Cache Update: Due to dynamic vehicle surroundings, the data features are periodic and fluctuate over time [6]. Age of Information (AoI) [7], which is defined as the elapsed time from data generation, has been extensively investigated as an effective metric for evaluating data freshness. The AoI, however, is still weak to differentiate between distinct degradation rates in heterogeneous data. Consequently, a new metric has to be designed to measure data freshness in a comprehensive way. Additionally, due to data temporality, a number of cache update methods [8], [9] are designed to regularly update data at edge node by leveraging the vehicles equipped with multiple sensors. In particular, data access preference is difficult to predict due to dynamic vehicular factors. Fortunately, researchers presented numerous learning-based cache update strategies [10] by exploiting deep learning models. However, most studies made the assumption that vehicles voluntarily assist with data upload in the absence of an appropriate incentive scheme.

2)

Blockchain Parameter Adaptation: blockchain technology, which is utilized as a distributed ledger for Bitcoin economic transactions, has been introduced in vehicular networks to guarantee data security and authenticity during V2X communication [11]. They applied blockchain technology in various vehicular applications, such as data trading and sharing [12], transportation service [13], authentication [14] and resource sharing [15]. These researchers have focused on developing various authentication techniques [16] and consensus mechanisms [17] in blockchain-based frameworks, in order to meet specific security requirements. Particularly, they have designed parameter optimization technique [18], block verifier selection method [19], or reputation mechanism [20] to assure the accuracy and reliability of data transmission and interaction. However, they have scarcely analyzed the procedure of the consensus mechanism, where the effect of parameters on consensus latency is not theoretically investigated.