Scholarlink

<p>The growth of the <i>Internet of Things</i> (IoT)-related innovations has resulted in the invention of numerous IoT objects. However, the resource limitations of individual items remain a challenge that can be overcome through offloading. A key limitation of previous research is the absence of an integrated offloading framework that can operate securely in offline/online environments. The security and calculated online/offline offloading issues in a multi-user IoT-fog-cloud system with blockchain are investigated in this article at the same time. First, we provide a reliable access control system utilizing blockchain to enhance offloading security. This technique can guard cloud resources against unauthorized offloading practices. Next, we define a computation offloading issue by optimizing the offloading decisions, allocating computing resources and radio bandwidth, and intelligent contract use to address the computation management of authorized mobile devices. This optimization challenge focuses on the long-term system costs of latency, energy use, and intelligent contract charge among all mobile devices. We create a new <i>Deep Reinforcement Learning (DRL)</i> technique employing a double-dueling Q-network to address the suggested offloading problem. We provide a <i>Markov Decision Process</i> (MDP)-based DRL solution to the IoT offloading-enabled blockchain dilemma. The supposed system works in both online and offline settings, and when operating online, we use the <i>Post Decision State</i> (PDS) method. The contributions of this work include a new integrated offloading framework that can operate in offline/online environments while preserving security and a novel approach that incorporates fog platforms into IoT blockchain-enabled networks for improved system efficiency. Our method outperforms four benchmarks in cost by 5.1%, computational overhead by 4.1%, energy use by 3.3%, task failure rate by 3.6%, and latency by 3.9% on average.</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

Securing and optimizing IoT offloading with blockchain and deep reinforcement learning in multi-user environments

  • 期刊论文

  • 发表日期：

    2025年3月14日

查看全文
  • Arash Heidari,
  • Nima Jafari Navimipour,
  • Mohammad Ali Jabraeil Jamali,
  • Shahin Akbarpour

摘要

The growth of the Internet of Things (IoT)-related innovations has resulted in the invention of numerous IoT objects. However, the resource limitations of individual items remain a challenge that can be overcome through offloading. A key limitation of previous research is the absence of an integrated offloading framework that can operate securely in offline/online environments. The security and calculated online/offline offloading issues in a multi-user IoT-fog-cloud system with blockchain are investigated in this article at the same time. First, we provide a reliable access control system utilizing blockchain to enhance offloading security. This technique can guard cloud resources against unauthorized offloading practices. Next, we define a computation offloading issue by optimizing the offloading decisions, allocating computing resources and radio bandwidth, and intelligent contract use to address the computation management of authorized mobile devices. This optimization challenge focuses on the long-term system costs of latency, energy use, and intelligent contract charge among all mobile devices. We create a new Deep Reinforcement Learning (DRL) technique employing a double-dueling Q-network to address the suggested offloading problem. We provide a Markov Decision Process (MDP)-based DRL solution to the IoT offloading-enabled blockchain dilemma. The supposed system works in both online and offline settings, and when operating online, we use the Post Decision State (PDS) method. The contributions of this work include a new integrated offloading framework that can operate in offline/online environments while preserving security and a novel approach that incorporates fog platforms into IoT blockchain-enabled networks for improved system efficiency. Our method outperforms four benchmarks in cost by 5.1%, computational overhead by 4.1%, energy use by 3.3%, task failure rate by 3.6%, and latency by 3.9% on average.