Search

Reputation evaluation is an effective measure for maintaining secure Internet of Things (IoT) ecosystems, but there are still several challenges when applied in IoT-enabled pumped storage power stations (PSPSs), such as the limited resources of intelligent inspection devices and the threat of single-point and collusion attacks. To address these challenges, in this paper we present ReIPS, a secure cloud-based reputation evaluation system designed to manage intelligent inspection devices' reputations in IoT-enabled PSPSs. Our ReIPS incorporates a resource-rich cloud platform to collect various reputation evaluation indexes and perform complex evaluation operations. To resist single-point attacks, we present a novel reputation evaluation model that combines backpropagation neural networks (BPNNs) with a point reputation-weighted directed network model (PR-WDNM). The BPNNs objectively evaluate device point reputations, which are further integrated into PR-WDNM to detect malicious devices and obtain corrective global reputations. To resist collusion attacks, we introduce a knowledge graph-based collusion device identification method that calculates behavioral and semantic similarities to accurately identify collusion devices. Simulation results show that our ReIPS outperforms existing systems regarding reputation evaluation performance, particularly in single-point and collusion attack scenarios. [ABSTRACT FROM AUTHOR]

Reputation evaluation is an effective measure for maintaining secure Internet of Things (IoT) ecosystems, but there are still several challenges when applied in IoT-enabled pumped storage power stations (PSPSs), such as the limited resources of intelligent inspection devices and the threat of single-point and collusion attacks. To address these challenges, in this paper we present ReIPS, a secure cloud-based reputation evaluation system designed to manage intelligent inspection devices’ reputations in IoT-enabled PSPSs. Our ReIPS incorporates a resource-rich cloud platform to collect various reputation evaluation indexes and perform complex evaluation operations. To resist single-point attacks, we present a novel reputation evaluation model that combines backpropagation neural networks (BPNNs) with a point reputation-weighted directed network model (PR-WDNM). The BPNNs objectively evaluate device point reputations, which are further integrated into PR-WDNM to detect malicious devices and obtain corrective global reputations. To resist collusion attacks, we introduce a knowledge graph-based collusion device identification method that calculates behavioral and semantic similarities to accurately identify collusion devices. Simulation results show that our ReIPS outperforms existing systems regarding reputation evaluation performance, particularly in single-point and collusion attack scenarios.