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.

In a big data system, infrastructure is made up like that large number of information is stored on a server which has all client's data and other data also and that data is used by users, basically they host the data. Information security is considered as a major challenge in such system. From a client's standpoint, the biggest risks in using big data systems is that they have to believe on the service provider of big data system, this system are owns and designed by service provider, user have to store and access that data so that they have lot of risk about it. Methodology: This work propose a new system architecture in which insider attacks can be identified by using the repetition of data on different nodes in the system. From all of the attacks, Insider attacks are one of today's most difficult cyber security problem that are not well addressed by commonly employed security solutions. Until several scientific research paper published in domain of insider attack, this paper certify that the field can benefit from the proposed structure, taxonomy and novel categorization of research that contribute to the organization of insider attack incidents and the defense solutions used for them. The target of our order is to systematize learning in insider threat research, while utilizing existing ground theory strategy for rigorous literature review Work process of proposed system categories among some classes that include: 1) Events and datasets, 2) Examination of attackers, 3) Process act, and 4) Defense solutions. Special attention is paid to the definitions and taxonomies of the insider threat; we present an auxiliary scientific classification of insider threat incidents, which is based on existing taxonomies. Outcome: This paper will help to improve researcher's work in the domain of insider attack, because it provides following things: 1) Time-to-time an updated and mostly available datasets that can be helpful for testing new detectionsolutions against different attack, 2) References of existing case studies and architecture of insider's behaviors is used for the purpose of testing defense solutions or expanding their coverage, 3) An exchange of knowledge about current patterns and further research directions that can be used for thinking in the insider risk space. [ABSTRACT FROM AUTHOR]