Trust Mechanism of Feedback Trust Weight in Multimedia Network

Trust mechanism was originally used in social science, psychology, management, and other disciplines. In the current multimedia network era, the trust mechanism has attracted many scholars’ attention since it enters computer science and network. The successful application of eBay's trust model [1] in eBay system (TMBS) makes scholars strongly interested in applying a trust mechanism in the network. Therefore, there have been many studies on the computer network's security and trust mechanism.

Graves et al. [2016] proposed a machine learning model called the DNC (differentiable neural computer). When trained with supervised learning, they demonstrated that the DNC could successfully answer synthetic questions designed to emulate reasoning and inference problems in natural language [2]. Lakshmi et al. [2017] proposed an HTTP GET Flooding Detection and Confidence-Based filtering method for DDoS Attack Defense in Web application to solve the Distributed Denial of Service (DDoS) attack in Web-based and Client-Server applications. They also applied dynamic resource allocation to automatically coordinate the available resources (CPU, memory, I/O, and bandwidth) of the server to mitigate DDoS attacks on individual clients [3]. Yuan et al. [2018] proposed a reliable and lightweight trust mechanism for IoT edge devices based on multi-source feedback information fusion to increase IoT edge computing applications’ adoption rate. Finally, they found that the mechanism overcame traditional trust schemes’ limitations in the artificial or subjective weighting of trust factors [4]. Zhang et al. [2019] proposed using reputation value to represent the trust degree of nodes to collect valuable information. They put forward a reputation management system composed of reputation information collection and reputation value calculation, advocated a new trust collection mechanism to optimize trust information collection efficiency, and designed a subjective trust calculation model based on trust degree. Finally, they verified the effectiveness of the model by simulation [5].

Besides, the rising social media research is just unfolding and has paid little attention to the topic of how to win users’ trust and establish social media authority so far. Magoi et al. [2020] focused on implementing the trust mechanism of social media in the library. The results showed that social media authority was established based on user trust. Based on this observation, it was found that the successful implementation of social media in libraries was related to librarians’ proficiency in using social media and how they constructed their social media content through trust-building activities [6]. As the trust mechanism is widely applied, the traditional trust model fails to adapt to the increasingly complex application environment. Abundant trust mechanisms based on P2P networks have a single evaluation basis, which cannot meet the needs of emerging applications such as e-commerce and social networks. Therefore, it is necessary to consider application scenarios and context. In this way, a context-based trust mechanism has become a research hotspot, and an FTWDF is proposed. The algorithm evaluates nodes’ status by a feedback mechanism of data trust and communication trust within clusters and evaluates cluster heads’ status by a feedback mechanism of communication trust between clusters. This algorithm ensures the data reliability of the node and improves the accuracy of data fusion.

In a multimedia network, wireless sensor nodes can collect and process data information through wireless technology to provide helpful information to users. However, in practical applications, WSN often receives interference from the external environment, which makes sensor nodes untrusted. Therefore, it is essential to build a trust mechanism in networks to prevent it from attack from unknown networks.

Prabhu et al. [2017] discussed the available scenarios of using sensor nodes in military applications and analyzed the nodes of trust mechanism based on the basic standards that must be considered in applying and deploying wireless sensors. Finally, according to the distance between the sensor node and the target, they located the intruder from latitude and vertical coordinates [7]. Jiang et al. [2017] studied the trust mechanism of UWSNs (underwater wireless sensor networks), proposed a trust cloud model suitable for trust management of UWSNs, and verified its effectiveness [8]. Yi et al. [2018] studied the coverage vulnerability in WSNs. Additionally, to offset the negative impact of coverage vulnerability, they studied the credible information coverage (CIC) vulnerability detection, namely CICHD, based on the CIC model. Finally, they found that the constructed algorithm could effectively detect the location and number of coverage holes [9]. Guo et al. [2019] used the perception transfer theory and the viewpoint of privacy calculus to explore the internal mechanism of users’ behavior response to help seeking activities. Experiments showed that privacy protection and intimacy affected the recipients’ privacy concerns and social rewards, affecting their participation behavior [10]. Mohiuddin et al. [2020] proposed a fuzzy-based trust management mechanism (FTM-IoMT) to provide trust management for electronic health systems users using IoMT infrastructure. Finally, it is found that the mechanism is helpful for IoMT nodes to collect authentic information from neighboring nodes while ignoring Sybil nodes, showing superior results [11].

The analysis of the above findings suggests that in WSNs, the data collected between neighbor nodes will produce redundant data. Therefore, data fusion technology is used to reduce redundant data to improve the accuracy of data fusion results and reduce the energy consumption in WSNs. However, these functions of data fusion technology are achieved based on increasing the performance consumption. First, it increases the average delay of the network. In the process of data transmission, priority will be given to the routing path with data fusion, or data fusion will be carried out directly in the network or after all nodes have transmitted data in a certain period. Second, it reduces the robustness of the network. Data fusion technology reduces the total amount of data transmission and loses a lot of data information, which increases the invalid data and data loss of network nodes. In environmental monitoring, due to the complexity of the deployment environment and the lack of secure communication mode, there are untrusted nodes in sensor networks; that is, the nodes are not reliable. The wrong behavior of unreliable nodes may reduce the accuracy and reliability of the data collected by the nodes, which directly affects the user's demand analysis and decision-making. Therefore, it is vital to identify the reliability of member nodes in the network.

The application of recommendation algorithm is as critical as the study of the trust mechanism in multimedia networks. At present, the recommendation algorithm in the multimedia network can be divided into two steps. Step 1: Learning. Establish a heuristic model based on the historical behavior data of users. Step 2: Applying. Recommend according to the heuristic model to predict the user's behavior preferences.

Chiregi et al. [2017] discussed the trust evaluation mechanism in cloud computing and divided it into two categories, including centralized and distributed. Meanwhile, they defined the integrity, security, availability, reliability, security, dynamic, confidentiality, and scalability of trust, and discussed the trust application and recommendation algorithm, including monitoring and tracking [12]. Wang et al. [2018] proposed a credible aware model (ACOSR) based ant colony optimization algorithm for secure routing in WSNs to resist malicious nodes’ internal attacks. Simulation results indicated that the algorithm's performance was significantly improved in packet loss rate, end-to-end delay, throughput, and energy consumption [13]. Liu et al. [2019] applied a trust mechanism to the “Internet+” production and research innovation, recommended an algorithm for recommending network education mode, and analyzed its potential security problems. They found that there were still many challenges in privacy protection [14].

There are many malicious and counterfeit web pages in the multimedia network; hence, it is necessary to retrieve trusted web pages. Personalization of web page recommendations is necessary to evaluate users’ interest in recommending web pages according to their own choices. Deepak et al. [2020] proposed a hybrid strategy-based personalized Web recommendation framework. According to the user's query results, web pages were recommended by analyzing the user's Web usage data. A series of strategies were intelligently integrated to achieve an efficient web page recommendation system [15]. With scholars’ vigorous research in related fields, the collaborative filtering algorithm has become one of the most accurate and practical methods in all recommendation technologies. Many scholars are studying this technology and applying it in industrial application systems. The recommendation algorithm assumes that users will still like the items they liked before to make a personalized recommendation for them through historical information such as past purchases. Usually, this method first analyzes the interaction between users and items that have been collected and then uses these interactions to generate items recommendation for users. The recommendation model based on normalized matrix factorization is improved in the accuracy and scalability of recommendation. However, for the massive amount of data in reality, the model's characteristics based on serial training seriously affect the parallelization of iteration and further affect the real-time performance of recommendation. Therefore, it is necessary and significant to realize the parallelization of recommendation algorithm models, especially with the continuous improvement of computer hardware performance, the widespread popularity of computer clusters, and the increasingly perfect parallelization platform and technology.

Fuzzy logic usually refers to the fact that things can express fuzzy uncertain states except absolute “true” and “false” in logic language, such as “unknown state” or “fuzzy uncertainty.” Many researchers have studied and applied fuzzy logic in the development of the trust mechanism of multimedia networks.

Asha et al. [2017] proposed a routing algorithm based on fuzzy clustering to solve the security and energy consumption problems in WSNs. They finally found that the model could quickly detect and obtain node trust, to improve the security of data routing [16]. The message security transmission strategy in the opportunity network (OppNets) is a complex and unique problem, entirely different from the traditional network using TCP/IP protocol [17]. Chhabra et al. [2018] proposed a security protocol FuzzyPT (potential thread), for blackhole attacks in OppNets, and applied the evolutionary game theory model to analyze the relay's decision-making ability to choose different strategies when forwarding messages. Finally, they found that the proposed protocol was superior to the existing security protocols based on cognitive and evolutionary game theory [18]. The social network based on a multi-agent model can store its environment's information and share it with other system agents. It is of great significance to implement a secure communication setting and select the appropriate receiver according to the information requested. Combining social relations, Hussain et al. [2019] proposed a trust-based fuzzy reasoning model in the multi-agent system. They analyzed the influence, credibility, and risk of cognitive agents. The model was implemented by Dempster-Shafer theory (DST), and the results were compared [19]. Sumalatha et al. [2020] proposed a cross-layer security-based fuzzy trust computing mechanism (CLS-FTCM). Compared with previous technologies, CLS-FTCM was an efficient wireless sensor environment security technology with high detection accuracy [20].

The above scholars’ research reveals that, although there are many studies on trust mechanism in related fields, there is no unified advantage in applying various algorithms. Most of them use one algorithm and have no prominent advantage. Therefore, recommendation algorithms, fuzzy logic theory, and wireless sensor nodes are collaboratively used in the trust mechanism of multimedia networks to solve security loopholes, malicious nodes, and energy consumption in multimedia networks. Based on the FTWDF algorithm, the fuzzy logic method is used to explore the “hot zone” problem and energy balance problem, combined with a new non-uniform clustering algorithm EEFA. In this algorithm, fuzzy logic is applied to cluster head selection and node clustering. First, node degree and residual energy are considered to generate candidate cluster head nodes. Then, considering the residual energy, node degree, and the distance from the node to the base station, the fuzzy logic algorithm is used to estimate the probability of each candidate cluster head being selected as the cluster head and the maximum node capability of the cluster head node. The node with the more significant probability value is regarded as the final cluster head and then cluster according to the maximum node capability of the cluster head. In the node clustering stage, ordinary nodes join the nearest cluster head until all nodes are clustered. In conclusion, the FTWDF-EEFA clustering algorithm provides a reference for improving the security of the multimedia network.