**2.6. Trust and reputation**

In social networks, Trust and Reputation are generally the two important components which play a major role in establishing relationship between entities which have been studied mainly by social scientists for a long time. All kinds of daily transactions, interactions, and communications in human life are based on trust. In a human social community, trust between two individuals is developed based on their actions over time. When faced with uncertainty, individuals trust and rely on the actions and opinions of others who have behaved well in the past. When affairs are to be handled in social networks, people always consider trust and reputation of concerned parties as prime tools for decision making.

Trust in general is the level of confidence in a person or a thing. More precisely trust can be defined as: "the quantified belief by a trustor with respect to the competence, honesty, security and dependability of a trustee within a specified context" [36, 37]. Reputation is a notion sometimes confused with trust; it is defined as "the global perception about the entity's behavior norms based on the trust that other entities hold in the entity" [38]. Reputation is the opinion of one person about the other, of one internet buyer about an internet seller, and one WSN node about another. Trust is a derivation of the reputation of an entity. Based on a reputation, a level of trust is bestowed upon an entity. The reputation itself has been built over time based on that entity's history of behaviour, and may be reflecting a positive or negative assessment.

In Wireless Sensor Network routing approaches, Reputation system based trust models borrowed from human societies have been proposed to combat misbehaviors. Nodes establish trust relationships between each other and base their routing decisions not only on geographical or pure routing information, but also on their trust that their neighbors will sincerely cooperate. In the context of WSN, Trust is the confidence of one node on another node that it will perform the given task as expected with full cooperation without any deviation. To evaluate the trustworthiness of its neighbors, a node not only monitors their behavior, i.e., through direct observations also known as First Hand Information(FHI), but may also communicate with other nodes to exchange their opinions , i.e., through indirect observations also known as Second Hand Information(SHI). The methods for obtaining trust information and defining each node's trustworthiness are referred to as trust models. A trust model is mostly used not only for higher layer decisions such as routing [39,40], data aggregation [41], but also for cluster head election [42] and for key distribution [43]. Goal of the trust model is to improve security thereby increasing the throughput, the lifetime and the resilience of a wireless sensor network.

242 Wireless Sensor Networks – Technology and Protocols

for many aspects in WSN.

conditions.

**2.5. Routing attacks** 

others can be found in [35].

**2.6. Trust and reputation** 

for decision making.

reflecting a positive or negative assessment.

consideration in WSN compared to other types of ad hoc network in the sense that WSN when deployed it is mainly focused on how to satisfy the environmental

 *Application specifications*: While normal ad hoc networks can be usually thought as general purpose networks, the whole WSN is built to serve a specific application. Therefore, WSN must satisfy the application requirements in addition to the environment conditions. This complicates the issue of finding general purpose solutions

There are several attacks that target the network layer in WSN. For example, in the blackhole attack, adversary nodes do not forward packets completely, while it selectively forwards some packets in gray-hole attack. Another example is the sybil attack in which a node pretends multiple identities. Thus, such a node can virtually exist in different neighborhoods and drop more packets. Wormhole attack is a collusion based attack in which an agreement between two adversaries is made to perform other attacks like blackhole. In wormholes, one adversary misroutes a received packet and sends it to its partner by faking a good routing decision. A detailed explanation of these attacks and

In social networks, Trust and Reputation are generally the two important components which play a major role in establishing relationship between entities which have been studied mainly by social scientists for a long time. All kinds of daily transactions, interactions, and communications in human life are based on trust. In a human social community, trust between two individuals is developed based on their actions over time. When faced with uncertainty, individuals trust and rely on the actions and opinions of others who have behaved well in the past. When affairs are to be handled in social networks, people always consider trust and reputation of concerned parties as prime tools

Trust in general is the level of confidence in a person or a thing. More precisely trust can be defined as: "the quantified belief by a trustor with respect to the competence, honesty, security and dependability of a trustee within a specified context" [36, 37]. Reputation is a notion sometimes confused with trust; it is defined as "the global perception about the entity's behavior norms based on the trust that other entities hold in the entity" [38]. Reputation is the opinion of one person about the other, of one internet buyer about an internet seller, and one WSN node about another. Trust is a derivation of the reputation of an entity. Based on a reputation, a level of trust is bestowed upon an entity. The reputation itself has been built over time based on that entity's history of behaviour, and may be Trust in WSN plays an important role in constructing the network and making the addition or deletion of sensor nodes from a network very smooth and transparent. Trust in WSN has been studied lightly by current researchers and is still an open and challenging field. Trust is an old yet important issue in any networked environment that can solve some problems which lies beyond the power of traditional cryptographic security. A Trust Management System is required to support the decision making processes of the network. Trust management is fundamental to identify malicious, selfish and compromised nodes which have been authenticated. In wireless sensor network, trust management system aids the nodes termed as trustors to deal with uncertainty about the future actions of other participating nodes termed as trustees. By evaluating and storing the reputation of other members, it is possible to calculate how much those members can be trusted to perform a particular task. It has been widely studied in many network environments such as peer-topeer networks, grid and pervasive computing and so on. However, in reality, sensor nodes have limited resources and other special characters, which make trust management for WSNs more significant and challenging. Various Trust models, Trust evaluation metrics and Trust Management schemes have been reported in the literature[36-59]. Current research on the trust management mechanisms of WSNs have mainly focused on nodes' trust evaluation to enhance the security and robustness. The practical applications of this method include the route, data integration and cluster head vote[44]. Although some existing approaches have played greater roles in improving security of other ad-hoc networks, trust management in WSNs still remains a challenging field.

The trust problem is a *decision problem under uncertainty*, and the only coherent way to deal with uncertainty is through *Probability*. There are several frameworks for reasoning under uncertainty, but it is well accepted that the probabilistic paradigm is the theoretically sound framework for solving decision problems with uncertainty. Some of the trust models introduced for sensor networks employ probabilistic solutions mixed with ad-hoc approaches. The problem of assessing a reputation based on observed data is a statistical problem. Some trust models make use of this observation and introduce probabilistic modeling that uses a Bayesian updating scheme known as the Beta Reputation System [65] for assessing and updating the nodes reputations. The use of the Beta distribution is due to the binary form of the events considered. For example, RFSN[2] uses a probability model in the form of a reputation system to summarize the observed information (FHI) and share the values of the parameters of the probability distributions as second-hand information(SHI). This shared information is soft data, requiring a proper way to incorporate it with the observed data into the trust model. The step of combining both sources of information is handled differently by different trust models. RFSN uses Dempster-Shafer belief theory model [48], solving it using the concept of belief discounting, and doing a reverse mapping from belief theory to continuous probability. In [49], a new Bayesian fusion algorithm to combine more than one trust component - data trust and communication trust to infer the overall trust between nodes is proposed. The trust value calculated between nodes based on their cooperation in routing messages to other nodes in the network is termed as Communication trust (CT). The trust value calculated based on the actual sensed data of the sensors in WSNs is known as Data trust (DT). As an extension to this work, authors proposed Recursive Bayesian Approach to Trust Management (RBTMWSN)[50] by introducing a new trust model and a Gaussian reputation system(GRSSN) for wireless sensor networks based on a sensed continuous data. In this work, Bayesian probabilistic approach based on the work done in modelling Expert Opinion[51] for mixing second-hand information from neighboring nodes with directly observed information is proposed. Opinions provided by knowledgeable sources are known as experts opinions. Such opinions are modulated by existing knowledge about the experts themselves, to provide a calibrated answer. It allows for the formal incorporation of informed knowledge into a statistical analysis. The probabilistic approach adopted is to consider the opinion given by the expert as soft data that is merged with the hard data according to the laws of probability[52]. In [53], authors proposed a Node Behavioral Strategies Banding Belief Theory of the Trust Evaluation (NBBTE) Algorithm. In this approach, at first, each node establishes the direct and indirect trust values of neighbor nodes by comprehensively considering various trust factors such as packet receive, send, strictness, delivery, consistency and availability, *etc*, and combining these factors together with network security grade, correlation of context time and rewards degree. Next, fuzzy set theory is used to decide the trustworthiness levels in accordance with the fuzzy subset grade of membership functions. Based on the levels of trustworthiness, the basic confidence function of D-S evidence theory[54] is accordingly formed. Finally, using the revised Dempster rules of combination, the integrated trust value of a node is obtained by integrating its trustworthiness of multiple neighbor nodes.

Current research challenge has been in designing an accurate and efficient trust and/or reputation model for distributed and heterogeneous environments[47]. When developing such models, different issues have to be taken into consideration. The problem to be solved here consists of deciding in a distributed environment which entity is the most reliable to interact with, in terms of trust and reputation. That is, having a system where different entities offer some services or goods and other ones are requesting those services, the former will always look for the best self profit, while the latter will demand the best services with respect to some quality characteristics, properties or attributes. Nevertheless, most of the times it is not feasible or realistic to assume the existence of service level agreements or the presence of a centralized entity or architecture supplying reliable information regarding the actual and current behavior of every service provider in the system. Hence, requesters have to determine on their own which service providers are the best ones according to certain criteria. Under these conditions, trust and/or reputation models are aimed to select the most trustworthy entity all over the system offering a certain service.
