«Opportunistic IoT: Exploring the Harmonious Interaction between Human and the Internet of Things Bin Guo1 Daqing Zhang1,2 Zhu Wang1 Zhiwen Yu1 ...»
We thus need to exploit various social behaviors in designing broker-based data dissemination protocols. For example, when selecting brokers, the social features such as user popularity (does the broker meet many different people each day), social willingness (is the broker willing to carry and forward the message), social network structure (Bob’s friends are more likely to act as his brokers), preferences (the broker may filter dissemination it tasks according to his/her preferences), and so on, will affect the performance of the protocol designed. Therefore, we state that the application of opportunistic IoT is also driven by exploring human behaviors and social features. One use case is illustrated in Fig. 2, where a broker is selected to carry and forward a message in the school campus, by measuring her social features such as social popularity and willingness.
Figure 2: The impact of human behaviors on the opportunistic IoT Besides data routing, there are several other human factors that may affect the formation and performance of opportunistic IoT. For example, human usually carry different kinds of mobile devices (mobile phones, PDAs, etc.), with distinct capabilities (some are more powerful with respect to the other nodes). We thus should consider device heterogeneity when designing the networking protocols for opportunistic IoT. In decentralized environments such as opportunistic IoT, trust relationship among peers also plays an important role. As mentioned earlier, the development of services in opportunistic IoT (e.g., sharing local-sensed information; formation of opportunistic communities for common-goal achieving) relies on the collaboration among opportunistically-encountered people. Mechanisms for establishing trust are thus crucial to maintain information security and data privacy in opportunistic communities. We discuss about this in detail in Section 4.3.
Overall, the bi-directional effects between human and IoT reflect the basic nature of opportunistic IoT. It also reveals the social side (while not technical side) of IoT and presents the human-centric (while not thing-oriented) view of IoT, which has been little concerned in previous studies of IoT.
4. Challenges and Research Opportunities Developing the potential benefits offered by opportunistic IoT poses a number of challenges and concerns. In facilitating the development of opportunistic IoT systems, a fundamental issue is the design of data dissemination protocols. Other important issues include heterogeneous community orchestration, security and incentive mechanisms for user collaboration, and so on.
4.1. Human Behavior and Data Dissemination Data dissemination in opportunistic IoT is a difficult problem. The heuristic behind the dissemination policy is that, since content providers and content consumers might be completely unaware of each other in a dynamic network, and never be connected at the same time to the same part of the network . Therefore, data objects should be moved and replicated in the network in order to carry them to interested users despite disconnections and partitions.
As presented in Section 3.2, to facilitate data dissemination and reduce its cost, the broker-based solution is often used. To this end, researchers start to explore mobility models [17, 18], co-location patterns [19, 20], and social network structure  as key pieces of human behavior/context information to predict nodes’ activeness and estimate their “social popularity” to serve as brokers in the near future. This seems to be promising because contacts between nodes are fundamentally tied with human behaviors. Two basic assumptions are leveraged here: (1) The higher a node’s popularity, the higher the chances of it meet more devices; and (2) all users are willing to act as brokers (the so-called “selflessness brokers”). However, the latter assumption does not always hold, since brokers have to contribute computational resources during the data carrying and forwarding process. According to the social theories, socially selfish is a basic attribute of human beings [22, 23], which will affect human behaviors.
Besides, preferences will also affect the behaviors of a broker. Therefore, we should measure the affects of various social features and taking consideration of them when designing data dissemination protocols for opportunistic IoT systems.
4.2. Heterogeneous Community Orchestration With the development and prevalence of opportunistic communities, people will live in heterogeneous social communities within cyber-physical spaces - both online communities and social networks where digital content is exchanged, and in the physical world, which exploits opportunistic contacting (i.e., face-to-face) between pairs of networked devices (e.g., smart phones) to exchange each other’s content.
Different social networks have distinct features in terms of geographical coverage, infrastructure support, function time, and so on. This also leads to distinct human interaction patterns (e.g., comment/like in online communities, co-location in ad hoc communities) and implicit social knowledge (e.g., friendship/trustworthy/ influence in online communities, social popularity/movement patterns in ad hoc communities) that can be extracted from them. Study of the interaction between opportunistic and online social networks (e.g., how does online social network data mirror physical events), as well as merging their complementary features and fully combining their merits (e.g., connecting the two forms of social networks to enhance data dissemination/sharing), however, become an important yet challenging research area. We use the term “heterogeneous community orchestration” to represent the potential interaction/collaboration issues raised in multiple, heterogeneous, virtual/physical community environments.
So far, research on online and opportunistic communities follow two separate research lines. The interaction/collaboration of the two forms of communities has yet little been explored. There have been studies about social network analysis across heterogeneous networks. For example, Tang et al.  developed a framework for classifying the type of social relationships by learning across different networks (e.g., email network, mobile communication network). Researchers from CMU study the relationship between the users’ mobility patterns and structural properties of the online social network, to identify the implicit social link between physical interaction and online connection . Lee et al. proposed a geo-social event detection method by mining unusually crowed places (e.g., reporting social events such as festivals or protests) from geo-tagged Twitter posts . However, numerous open issues remain unexplored, such as the aggregated/collaborative effects of distinct social networks, data dissemination over heterogeneous social networks, and so on.
4.3. Security and Incentive Mechanisms for User Collaboration The sharing of data in opportunistic IoT applications can raise significant security concerns, with information being sensitive and vulnerable to privacy attacks. For example, in the opportunistic trading scenario, sensitive personal information such as user location, mobility patterns, preferences may be used by data dissemination protocols. The new security challenge introduced here is the protection of the privacy of participants while allowing their devices to reliably share/forward data in opportunistic IoTs. Data anonymization techniques , which conceal the identity of users when they contribute/forward data, can be one way to deal with this problem, but there are still many issues to be addressed in the future.
In opportunistic IoT, anonymous contributors are often used as brokers to carry and forward data. If there lacks the control over ensuring source validity and information accuracy, data credibility issues may arise. For example, the source node may send incorrect data; malicious nodes may modify the data it received and forward it to other nodes. Therefore, trust maintenance and abnormal detection methods should be built into opportunistic IoT systems to determine the trustworthiness and quality of the data being transmitted. However, traditional strategies often rely on online authentication from centric servers, which cannot meet the opportunistic connection and decentralized nature of opportunistic IoT systems. There are two possible ways to address this. First, we should follow a basic rule that the attack to a network is largely dependent on what kind of routing mechanism the opportunistic network uses. For instance, Uddin et al. have proposed the protection mechanisms for address spoofing in opportunistic networks under the Spray-andWait protocol . Second, it is beneficial to leverage the close tie between online and opportunistic communities in opportunistic IoT. For instance, the trust relationship established among people in online social networks can be leveraged to strengthen security protection in opportunistic communities.
Opportunistic IoT offers immense potential to consumers and service providers.
However, for these innovations to evolve from ideas to tangible products for the mass market, many commercial issues also require resolution. For example, in broker-based data dissemination protocols, brokers need to contribute their computational resources to other nodes. However, the fact is that most opportunistic IoT devices (e.g., mobile phones, wearable sensors) have limited resources, such as energy and storage capacity.
Therefore, the development of a solid economic model is highly important, and additional strategies for incentives and reputation for data contributors are needed (references are those explored in peer-to-peer systems  and ad-hoc networks ).
5. A Conceptual Framework To facilitate the development of opportunistic IoT, a generic system framework is essential. The framework should provide a set of mechanisms for dynamic network management, human behavior analysis, and information sharing among mobile nodes. It should address most of the issues mentioned in the previous subsections and provide a uniform interface for information distribution/access by various applications. We have proposed a conceptual framework for opportunistic IoT systems, as shown in Fig. 3. It can be a starting point to build opportunistic IoT applications with framework support.
The framework is maintained on IoT devices, where the following basic components are involved: the opportunistic network management (dynamic, intermittent connectivity), trust/security/privacy maintenance (e.g., data anonymization, malicious node detection, data access control, data quality enhancement), resource management (e.g., bandwidth, storage, computing, energy), social feature extraction (e.g., social network analysis, user preference learning, mobility pattern mining), incentive mechanisms for user collaboration, and the library of various data dissemination protocols (flooding, popularity-based broker selection, and so on). It should be noted that to enhance the linkage and interaction with other forms of networks, especially online social networks, the infrastructure also has a component for heterogeneous community/network orchestration (HCO). The HCO component is responsible for exchanging useful information and handling data floating among distinct networks.
6. Our Practice to Opportunistic IoT The human-centric nature of opportunistic IoT brings new potentials in many application areas. We make a summary of our ongoing work in the following and present our insights on how to address the challenges faced by opportunistic IoT.
6.1. Opportunistic Mobile Social Networking Forging social connections with others is the core of what makes us human.
Opportunistic social networking aims to improve social connectivity in physical communities by leveraging the information detected by smart devices. The SOCKER application we developed is such an example, which can build ad-hoc communities of like-minded people . For instance, if Harry wants to organize a basketball game at weekend in the university campus, he can post a request to SOCKER and recruit participants who are basketball fans and who live nearby. A broker-based mechanism is used by SOCKER to facilitate the dissemination of community-formation requests in the campus-wide environment. Finally, people who are socially- and physically-close to each other are opportunistically recruited to participate this activity. The concept of broker-based community creation is illustrated in Fig. 4. For each opportunistic community OCi in the figure, users in solid and dash circles represent present brokers and previous brokers, respectively, while users in solid rectangles are the matched community members (e.g., basketball fans).
Fig. 4. Community creation in SOCKER In Fig. 4, Harry initiates a community creation task tm and serves as the first broker.
Once Harry moves into a new opportunistic community, he will disseminate tm to the users encountered for match-making (each user keeps a list of her interests), and the matched users will be added to the community member list (e.g., uB is added in OC1).
Afterwards, broker election is carried out in this opportunistic community based on a specific broker selection strategy, and the “broker-switch” action will be performed once there is a more effective broker (e.g., uA is selected as the new broker for tm in OC1). The dissemination process terminates when i) the required number of participants is found (e.g., at least five matched users should be found to organize a basketball game), or ii) the pre-specified request dissemination time is expired. For instance, Harry hopes that the community can be created within three days. We define it as the community creation expiry time.