Scholarlink

<p>Data placement in next-generation networks identifies frequently requested content and proactively stores it in edge nodes during off-peak periods to reduce retrieval latency. However, existing cache placement and eviction policies mainly focus on prediction algorithms, thereby neglecting the critical role of storage constraints in cloud–edge environments. Given this, we propose a reactive caching strategy that combines recommender systems with blockchain-enabled access logging, ensuring that caching decisions are perfectly tailored to each authenticated end device. Specifically, we introduce a two-phase framework built on Graph Neural Networks (GNNs) and the Hyperledger Fabric platform. First, we proposed a hybrid approach, the ARBAC scheme, by combining the strengths of Attribute-Based Access Control (ABAC) and Role-Based Access Control (RBAC). Next, we integrated chaincodes (i.e. programs that consist of smart contracts) to support the distributed ledger, creating a permanent and tamper-resistant record. Second, GNN models are employed to predict popular content for each authorized IoT device. Empirical evaluation confirms the superiority of the proposed GraphSAGE model; by utilizing generalized aggregation functions and neighborhood sampling, it outperforms LSTM, CNN, GCN, and GAT, yielding improvements of up to approximately 20% in R², 18% in F1-score, and 17% in precision, recall, and NDCG, respectively. Crucially, by exploiting its consensus mechanism and on-chain identity management, the blockchain-based ARBAC model guarantees fine-grained query verification while optimizing the trade-off between security and edge content caching performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A two-level security-aware reactive caching for IoT-edge data centers using deep learning and hyperledger fabric blockchain-driven access control

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月9日

查看全文
  • Mbarek Marwan,
  • Hind Ait Temghart,
  • Abdelkarim Ait Temghart,
  • Mohamed Lazaar

摘要

Data placement in next-generation networks identifies frequently requested content and proactively stores it in edge nodes during off-peak periods to reduce retrieval latency. However, existing cache placement and eviction policies mainly focus on prediction algorithms, thereby neglecting the critical role of storage constraints in cloud–edge environments. Given this, we propose a reactive caching strategy that combines recommender systems with blockchain-enabled access logging, ensuring that caching decisions are perfectly tailored to each authenticated end device. Specifically, we introduce a two-phase framework built on Graph Neural Networks (GNNs) and the Hyperledger Fabric platform. First, we proposed a hybrid approach, the ARBAC scheme, by combining the strengths of Attribute-Based Access Control (ABAC) and Role-Based Access Control (RBAC). Next, we integrated chaincodes (i.e. programs that consist of smart contracts) to support the distributed ledger, creating a permanent and tamper-resistant record. Second, GNN models are employed to predict popular content for each authorized IoT device. Empirical evaluation confirms the superiority of the proposed GraphSAGE model; by utilizing generalized aggregation functions and neighborhood sampling, it outperforms LSTM, CNN, GCN, and GAT, yielding improvements of up to approximately 20% in R², 18% in F1-score, and 17% in precision, recall, and NDCG, respectively. Crucially, by exploiting its consensus mechanism and on-chain identity management, the blockchain-based ARBAC model guarantees fine-grained query verification while optimizing the trade-off between security and edge content caching performance.