The “internet of things” (IoT) concept is used to define or reference systems that rely on the autonomous communication of a group of physical objects. The applications areas of the IoT are numerous, including: smart homes, smart cities and industrial automation. IoT systems often provide great benefits to numerous industries and society as a whole. Trust management plays an important role in IoT for reliable data synthesis and mining, qualified services with context-awareness, and enhanced user privacy and information security. It helps people overcome perceptions of uncertainty and risk and engages in user acceptance and consumption on IoT services and applications. However, current literature still lacks a comprehensive study on trust management in IoT. In this paper, providing a survey on the properties and objectives of IoT trust management, and also provide a survey on the current literature advances towards trustworthy IoT. Many of the IoT systems and technologies are relatively novel, there are still many untapped applications areas, numerous challenges and issues that need to be improved.
INTRODUCTION
The rising model of the Internet of Things (IoT) builds upon the cooperative connectivity of smart objects, including radio frequency identification (RFID) tags, sensors, actuators, PDAs, smart phones, etc. with wide applicability [1, 8]. The existing work of IoT has been focusing on the architecture of IoT and the enabling technologies for seamless cooperation among smart objects [1, 7-9, 19, 20, 23]. In addition, researchers have developed important IoT application scenarios, such as e-health [3, 11], smart-home and smart-community [15]. As the building blocks of IoT, smart objects with heterogeneous
characteristics need cooperatively work together to accomplish the application tasks. Another characteristic of IoT is that most smart objects are human-carried or human related devices. Therefore, the social relationships among the device users must be taken into consideration during the design phase of IoT applications. Further, devices in IoT very often expose to public areas and communicate through wireless. Hence, IoT objects are vulnerable to malicious attacks [17]. In this paper, we propose a trust management protocol for IoT considering both malicious and socially uncooperative nodes, with the goal to enhance the security and increase the performance of IoT applications.
Trust management (TM) plays an important role in IoT for reliable data fusion and mining, qualified services with context-aware intelligence, and enhanced user privacy and information security. It helps people overcome perceptions of uncertainty and risk and engages user in acceptance and consumption on IoT services and applications. Trust is a complicated concept with regard to the confidence, certainty, and expectation on the consistency, reliability, safety, dependability, ability, and other characters of an entity. Reputation is a measure derived from direct or indirect knowledge or experiences on earlier interactions of entities and is used to assess the level of trust put into an entity. However, the IoT poses a number of new issues in terms of trust. Generally, an IoT system contains three layers: a physical perception layer that perceives physical environments and human social life, a network layer that transforms and processes perceived environment data and an application layer that offers context-aware intelligent services in a pervasive manner. Each layer is intrinsically connected with other layers through cyber-physical social characteristics. A trustworthy IoT system or service relies on not only reliable cooperation among layers, but also the performance of the whole system and each system layer with regard to security, privacy and other trust-related properties. Ensuring the trustworthiness of one IoT layer (e.g., network layer) does not imply that the trust of the whole system can be achieved. Unlike
other networking systems, new issues are raised in the area of IoT caused by its specific characteristics.
First, data collection trust is a crucial issue in IoT. If the collected huge volumes of data from the physical perception layer are not trustworthy enough, e.g., due to the damage or malicious input of some sensors, the IoT service quality will be greatly influenced and hard to be accepted by users even though the network layer trust and the application layer trust can be fully provided.
Second, data process trust should be ensured. Trustworthy data fusion and mining require efficient, accurate, secure, privacy-preserved, reliable and holographic data process and analysis in a holistic manner.
However, achieving all trust properties in IoT data process is an arduous task hard to fulfill. On the other hand, IoT services are based on data process, analysis and mining. This fact actually greatly interrupts user privacy. At the same time when the users enjoy advanced services they also need to disclose or have to share their personal data or privacy. Intelligently providing context-aware and personalized services and at the same time preserving user privacy to an expected level introduces a big challenge in current IoT research and practice. More specifically, due to the cyber-physical and social characteristics of IoT, how to provide trustworthy services through social computing is a hot but uneasy topic. In the literature, trust and reputation mechanisms
have been widely studied in various fields. However, current IoT research has not comprehensively investigated how to manage trust in IoT in a holistic manner. There is little work on the trust management for IoT.
A number of issues, such as big data trust in collection, process, mining and usage; user privacy preservation; trust relationship evaluation, evolution and enhancement; user-device trust interaction, etc. have not been extensively studied. IoT introduces additional challenges to offer ubiquitous and intelligent services with high qualification in practice, especially when user privacy and data trust should be seriously considered and stringently supported. In this paper, we study trust properties and propose the objectives of IoT trust management. We explore the literature towards trustworthy IoT in order to point out a number of open issues and challenges and suggest future research trends related to trust management. We further propose a research model in order to achieve comprehensive trust management in IoT and direct future research.
The paper is organized as follows. We give an overview of related influential works in Section 2. Section 3 gives an overview of IoT System of involving three layers. Section 4 gives an overview of the literature towards trustworthy IoT. We conclude the paper in Section 5.
Related Work
Establishing security communication channels based on trust and reputation models among sensor nodes is an important consideration when designing a secure routing solution in IoT/CPS.
ATRM [2] is an agent-based trust and reputation management scheme for WSNs where trust and reputation management is carried out locally with minimal overhead in terms of extra messages and time delay. However, since mobile agents are designed to travel over the entire network and run on remote nodes, they must be launched by trusted entities. An agent-based trust model for WSN is presented in [4] using a watchdog scheme to observe the behavior of nodes and broadcast their trust ratings. Sensor nodes receive the trust ratings from the agent nodes, which are responsible for monitoring the former and computing and broadcasting those trust ratings. In [14], a reputation-based scheme called DRBTS is proposed to provide a method by which beacon nodes, BN, can monitor each other and provide information so that sensor nodes, SN, can choose who to trust, and based on a quorum voting approach.
However, in order to trust a BN’s information, a sensor must get votes for its trustworthiness from at least half of their common neighbors. BTRM-WSN [18] is a bio-inspired trust and reputation model for WSN aimed to achieve to the most
trustworthy path leading to the most reputable node in a WSN offering a certain service. Each node must maintain a pheromone trace for each of its neighbors. CONFIDANT [19] is proposed to extend reactive routing protocols with a reputation-based system in order to isolate misbehaving nodes. Each node monitors the behaviors of its next hop neighbors.
Trust relationships and routing decisions are based on experienced, observed, or reported routing and forwarding behavior of other nodes. SORI [20] scheme is proposed to encourage packet forwarding and discipline selfish behavior. The reputation of a node is quantified by objective measures, and the propagation of reputation is efficiently secured by a one way-hash-chain-based authentication scheme. Watchdog and Path rater mechanisms [16], are just two extensions to the DSR algorithm. However, not all of the most known works take into account the strong restrictions about processing, storage or communication capabilities. Some of them rely on a watchdog mechanism with or without using a multi-agent system. IoT/CPS assumes that trillions of things which are used on a daily basis will eventually be connected to the Internet employing 6LoWPAN [5] protocol and provide intelligent service through cooperating with each other.
Most things have the following significant characteristics [6] [10], limited power capability, wireless receivers and
transmitters with limited range facing the use of multi-hop communication, mobility (things will move, possibly become disconnected) and violability (things may be switched on and off frequently).
All the above issues raise the need for the development of a novel management model, different from those being in use today. Based on the research of characteristics of IoT/CPS and in-depth understanding of ATRM [2], ATSN [4], DRBTS [16], BTRM-WSN [18], CONFIDANT [19], SORI [20] and WP [16], we propose a novel trust and reputation model TRM-IoT to enforce the cooperation between things in a network of IoT/CPS based on their behaviors.
System model of IoT
We consider an IoT system involving three layers, as illustrated in Fig. 1: a physical perception layer that contains a huge number of sensors, actuators, mobile terminals and sensor connectors and applies sensing technologies to sense physical objects (including human beings) and social environments by collecting huge amount of data in order to convert them into the entities in the cyber world; a network layer that includes all network components with heterogeneous network configurations (e.g., wireless sensor networks, ad hoc networks, cellular mobile networks and the Internet) for data coding, transmission, fusion, mining and analyzing at data processors in order to provide essential information to an
application layer that pervasively and intelligently offers expected services or applications to IoT end users. This system model is compatible with the reference architecture model proposed by EU FP7 IoT-A project, especially the IoT-A tree structure. Meanwhile, various cyber-physical social relationships exist crossing the above three layers, which can be explored and mined to offer advanced services for human-beings. IoT trust management is concerned with: collecting the information required to make a trust relationship decision; evaluating the criteria related to the trust relationship; monitoring and reevaluating existing trust relationships; as well as ensuring the dynamically changed trust relationships and automating the process in the IoT system.
Fig 1. A System model of IoT
Protocol for Trust management in IoT
Trust management is done by use of some protocol called as Trust Management
Protocol. For IoT Trust Management Protocol is distributed. Each node maintains its own trust assessment towards other nodes. For scalability, a node may just keep its trust evaluation towards a limited set of nodes which it is most interested in. The trust management protocol is encounter based as well as activity-based, meaning that the trust value is updated upon an encounter event or an interaction activity. Two nodes encountering each other or involved in a direct interaction activity can directly observe each other and update their trust assessments. They also exchange their trust evaluation results toward other nodes as recommendations. In this trust management protocol, a node maintains multiple trust properties in honesty, cooperativeness, and community-interest. The trust assessment of node i evaluating node j at time t is denoted by T_ij^X (t) where X = honesty, cooperativeness, or community-interest. The trust value T_ij^X (t) is a real number in the range of [0, 1] where 1 indicates complete trust, 0.5 ignorance, and 0 distrust. When node i encounters or directly interacts with another node k at time t, node i will update its trust assessment T_ij^X (t) as follows:
T_ij^X (t)={█((1-α) T_ij^X (t) (t-∆t)+αT_ij^(X,direct) (t),@if j==k; @(1-γ) T_ij^X (t) (t-∆t)+γT_ij^(X,recom) (t),@ if j!=k;)┤ —-(1)
Here, ∆t is the elapsed time since the last trust update. If the trustee node j is node k itself, node i will use its new trust assessment toward node j based on direct observations 〖(T〗_ij^(X,direct) (t)) and its old trust node j based on past experiences to update T_ij^X (t). A parameter α (0 ≤ α ≤ 1) is used here to weigh these two trust values and to consider trust decay over time, i.e., the decay of the old trust value and the contribution of the new trust value. A larger α means that trust evaluation will rely more on direct observations. Here, T_ij^(X,direct) (t) indicates node i’s trust value toward node j based on direct observations accumulated over the time period [0, t]. Below we describe how each trust component value T_ij^(X,direct) (t) can be obtained based on direct observations for the case in which node i and node j interacting or encountering each other within radio range:
T_ij^(honesty,direct) (t): This refers to belief node i that node j is honest based on node i ‘s direct observations toward node j. Node i estimates T_ij^(honesty,direct) (t) by keeping a count of suspicious dishonest experiences of node j which node i observed during [0,t] using a set of anomaly detection rules such as a high discrepancy in recommendation has been experienced, as well as interval, retransmission, repetition, and delay rules as in [21,22]. If the count exceeds a system-
defined threshold, node j is considered totally dishonest at time t, i.e., T_ij^(honesty,direct) (t)=0. Otherwise T_ij^(honesty,direct) (t) is computed by 1 minus the ratio of the count to the threshold. The hypothesis is that a compromised node must be dishonest. Consider non-zero false positive probability (Pfp) and false negative probability (Pfn) for such detection mechanism.
Similarly we can calculate the T_ij^(cooperativeness,direct) (t), T_ij^(Community-interest,direct) (t) which provides the degree of cooperativeness of node j as evaluated by node i based on direct observations over [0,t] and the degree of the common interest or similar capability of node j as evaluated by node i based on direct observations over [0,t] respectively.
On the other hand, if node j is not node k, then node i will not have direct observation on node j and will use its past experience T_ij^X (t-∆t) and recommendations from node k (T_ij^(X,recom) (t) where k is recommender) to update T_ij^X (t). The parameter γ is used here to weigh recommendations vs past experiences and to consider trust decay over time as follows:
-γ=(βT_ik^X (t) )/(1+ βT_ik^X (t) ) — (2)
Here introduced another parameter β ≥ 0 to specify the impact of “indirect
recommendations” on T_ij^X (t) such that the weight assigned to indirect recommendations is normalized to
βT_ik^X (t) relative to 1 assigned to past experiences. Essentially, the contribution of recommended trust increases proportionally as either T_ik^X (t) or β increases.
Conclusion and Future Work
The Internet has changed drastically the way we live, moving interactions between people at a virtual level in several contexts spanning from the professional life to social relationships. The IoT has the potential to add a new dimension to this process by enabling communications with and among smart objects, thus leading to the vision of ‘‘anytime, anywhere, anymedia, anything” communications. Thus, we observe that the IoT should be considered as part of the overall Internet of the future. We analyzed system model of IoT and trust management protocol of IoT. The protocol takes social relationships into account and advocates the use of three trust properties, honesty, cooperativeness, and community-interest to evaluate trust. The protocol is distributed and each node only updates trust towards others of its interest upon encounter or interaction events. The trust assessment is updated by both direct observations and indirect recommendations. We analyzed the effect of trust parameters (α and β) on trust evaluation. The results demonstrate that (1) using more new direct observations over pass information could increase the trust
assessment accuracy and trust convergence speed, and (2) using more indirect recommendations over pass information could increase the trust convergence speed but decrease the accuracy in the presence of false recommendation attacks from malicious nodes.
In the future, there is a possibility to develop trustee-based and mission-based trust management for IoT. There is requirement of dynamic trust management for IoT and explore new trust-based IoT applications that can adapt to changing environments such as malicious node population/activities