CityWatch - Wiley Online Library

32 downloads 83777 Views 15MB Size Report
1 School of Computer Science and Statistics, Trinity College Dublin, Ireland ..... the best practice advice to those in the areas affected by the flooding. She can ...
TRANSACTIONS ON EMERGING TELECOMMUNICATIONS TECHNOLOGIES Trans. Emerging Tel. Tech. 2014; 25:64–80 Published online in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/ett.2786

SPECIAL ISSUE - SMART CITIES

CityWatch: exploiting sensor data to manage cities better Atif Manzoor1*, Constantinos Patsakis1 , Alistair Morris1 , Jessica McCarthy1,2 , Gabriel Mullarkey2 , Han Pham2 , Siobhán Clarke1 , Vinny Cahill1 and Mélanie Bouroche1 1 School of Computer Science and Statistics, Trinity College Dublin, Ireland 2 Intel Sustainable Cities Lab, Dublin, Ireland

ABSTRACT Persistent urbanisation of our planet places a continuous strain on cities’ resources and the quality of service delivery. While increasing city infrastructure might help alleviate this problem, the scale and complexity of future cities mean that this approach is unsustainable. Cities, however, are becoming increasingly instrumented with a myriad of sensors, both fixed and mobile. While a number of systems aim to exploit such sensors to gather information and to provide a real-time view of the city, existing approaches are application-specific, hindering their scalability and reuse. Using the city of Dublin (Ireland) as a testbed, this paper describes our iterative consultation process with city stakeholders to design CityWatch, an urban-scale data sensing and dissemination framework. In particular, it presents the resulting design of two prototype applications, the requirements on the overall framework, an initial implementation, and discusses the early results of ongoing trials. Copyright © 2014 John Wiley & Sons, Ltd. *Correspondence A. Manzoor, School of Computer Science and Statistics, O’Reilly Institute, Trinity College, Dublin 2, Ireland. E-mail: [email protected] Received 3 June 2013; Revised 30 November 2013; Accepted 1 December 2013

1. INTRODUCTION The 20th century phenomenon of worldwide urbanisation is persisting, and it is expected that around 60% of the world population will be living in cities by the end of 2030 [1]. This trend is creating increasing strains on cities’ resources and infrastructure, such as transportation and communication means, water and energy management and security [2]. Conditions further deteriorate during disruptive events or natural disasters, such as extreme weather conditions such as pluvial flooding and heavy snowfalls. In these circumstances, the spontaneous and real-time realisation of the situation is of pivotal importance to enable citizens, urban service providers and city management to cope with any ensuing disruption and to minimise their impact on citizens’ everyday life [3]. However, the traditional information collection methods used by city management councils and rescue authorities, such as phone calls and individual reports, are quickly saturated during disruptive events, for example, there may not be enough telephone lines. The scale, complexity and rate of change means that it is particularly challenging for humans to handle, gather and process the reported information to rapidly comprehend the current situation in the city. 64

Recent advances in information and communication technology, such as the deployment of the fourth generation of the mobile phone communication standards, miniaturisation of the computing devices and ubiquitous computing have resulted in our cities becoming instrumented with million of sensors [4, 5]. Current sensor systems are however typically single-purpose and disconnected. Our hypothesis is that gathering real-time information produced by such disparate existing systems can improve the management of city resources. Such systems, however, may be controlled by individual citizens, such as sensors embedded in smart phones; owned by private companies, such as closed-circuit television cameras coupled with dedicated fixed sensors; deployed by local authorities and utility companies, such as traffic detectors and smart energy and water metres or publicly shared, as seen in social media [6]. Advancements in urban and community sensing techniques, such as participatory sensing and opportunistic sensing, can take advantage of the finely grained instrumentation of our cities and capture big data [7]. But practical examples of such concepts are still rare. We believe that sustainable citizen participation and local authority engagement can only be achieved if city stakeholders are an integral part of the design process, from Copyright © 2014 John Wiley & Sons, Ltd.

A. Manzoor et al.

its very beginning. For this reason, we adopted a user-led design process, as an example of quadruple helix innovation [8] in action, where academic and industrial scientists, citizens and local government co-design applications to address societal challenges. To address our hypothesis, we have designed CityWatch (CW), an urban-scale data sensing and dissemination framework to enable the integration of disparate urban sensing systems, including individually owned data through participatory sensing. While the motivations for the CW data sensing and dissemination framework, an initial implementation and early results were presented before [9], this article puts these findings in context by presenting the design process and its outputs in details, discussing the disruptive events use case and demonstrating how the userled design process has informed the CW implementation, leading to a successful trial. In particular, this article first presents the user-led CW design process (Section 2) and the two resulting use cases, FloodWatch, which addresses real-time data gathering during extreme events (Section 3) and GreenWatch, which focuses on encouraging citizens to use the application in their daily life through an environmental use case (Section 4). Section 5 presents the overall outputs of the qualitative research in terms of requirements on the framework in particular. Subsequently, Sections 6 and 7 present the architecture and implementation of the CW and the GreenWatch use case, and Section 8 discusses early results of its ongoing trial. We also present a brief overview of the current state of the art research in the domain of smart cities and urban sensing applications (Section 9) and finally conclude the article (Section 10).

In particular, we used the qualitative data collected to construct CW usage narratives or scenarios. First, we synthesised the potential CW user goals and behaviours in the form of models, known as persona [10]. These persona are based on observations logged during the stakeholder interviews and increase CW stakeholder empathy and agreement within the design team by visualising the audience and defining goals [11]. When all persona were sufficiently characterised, we selected the primary persona as the target audience for the system. Afterwards, we used the primary persona to create CW usage scenarios to express the concept to stakeholders. While the usage scenarios themselves may comprise only a narrow spectrum of online possibilities, the vision of the project is to demonstrate informed possibilities for usage by multiple stakeholders across both their online and offline systems, including proposed derivative actions, behaviours and system management opportunities, hence validating our hypothesis. The scenarios draw attention to the high-level interactions and touch points between the product and the user. Importantly, they underline the functions of the product in order for the persona to accomplish their goals and not specifically how the product performs these functions. The final stage of the design research was the stakeholder workshop during which the team highlighted how CW could enhance visibility, meaning and action-ability of data in real-time in front of interested parties. The team found strong support for continued user-led innovation, which stakeholders felt could increase presence, develop partnerships, build on open data initiatives and strengthen future city planning. The following sections describe the use cases that were explored during the co-design process.

2. DESIGN PHILOSOPHY 3. FLOODWATCH Design research is a crucial element in the understanding of an innovation’s ecosystem. To test our hypothesis that gathering existing urban data could help manage cities better, we prepared an exploratory concept. To ground this concept in real-world stakeholder behaviours, actions and decision-making processes to inform the first technical prototype, the CW design team focused on qualitative research techniques, mainly stakeholder and subject matter expert interviews, field observations and a user-led design workshop. The activity flow of the CW design process is presented in Figure 1. Our intention in first understanding the group of stakeholders was to design not only for the technical, but also political and hierarchical, challenges of creating a rich data resource and decision-making asset for city management. This decision led to the ability to validate and improve upon our concept, build support and insight from different angles (from engineers in charge of resource modelling to the supervisors in charge of managing the front lines of the emergency call centre to how information is communicated to the press and thereby its citizens) and present these insights at a stakeholder workshop. Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

Because of limited time and field resource constraints in which there was less than 2 weeks to negotiate access and conduct fieldwork, the CW design team chose to focus in the first instance on a disruptive scenario, and in particular, on a singular use case that united the interests of diverse stakeholders within the city: the pluvial flood in Dublin in October 2011. This section presents the motivation for this use case, the approach to the design process, the personas derived from it, and finally discusses its results. 3.1. Motivation On 23 and 24 October 2011, Dublin weathered a pluvial flood event having the probability between 1% and 2% within 100 years. The over-topping and surcharging of flood management assets led to a complex flooding scenario across the city, resulting in 10.5 mega tonnes of water flooding the city within the span of 2 days. Dublin City Council made the efficient and timely decision to retain staff through the evening of 24 October to perform rescue activities. Emergency response plans were 65

A. Manzoor et al.

Figure 1. Activities performed during the CityWatch user-driven design process.

also implemented as soon as the risk of flooding became apparent. However, because of the intensity of the rainfall during the later part of the event and the subsequent city wide flooding, resources were stretched, and the communication overloaded. In particular, the city’s 200-year-old drains were not designed for both city urbanisation and facilitation of water flows of such intensity. Considering his experience to coordinate rescue activities during the aforementioned event, Tom Leahy, DCC Executive Manager, indicates ‘The idea that past events are a reliable guide to future events no longer applies due to climate change. Extreme events are now much more frequent requiring new adaptive strategies.’

These adaptive strategies—technological and social— must meet challenges posed by urbanisation, aged physical infrastructure and climate change. 3.2. Approach Considering the sensitivity and high profile of this event, the complexity of the response scenario by the city and other service providers and the need to not only vali66

date but also to gain support for CW as not only a data collection resource but also one that could be integrated into decision-making, we chose to structure our design process to first focus on understanding the city and serviceproviders’ experiences, viewpoints and insights stemming from the monster floods of 2011 resulting in more than 25 interviews with stakeholders including Dublin City Council, Dublin Flood Resilient Cities and Dublin Bus. This research (from fieldwork, workshops, to persona design) focused on understanding which types of data from participating sensing methods could be used, how they might integrate or strengthen with existing data or decision making resources, how citizens currently report information to city networks and what information may be missing, how the context of a disruptive event may affect the ability for citizens to provide information (or for the city to be able to collect this information) and what are the operational and decision-making hierarchies that will affect what information is required, trusted and how it needs to be filtered. Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

(a) Citizen using CityWatch

(b) Citizen management using CityWatch

Figure 2. Citizens and city management taking advantage of CityWatch in an urban environment: (a) citizen using CityWatch, (b) city management using CityWatch.

3.3. Personas and scenarios

3.4. Outcomes

The first scenario, as shown in Figure 2(a), focuses on a citizen user interacting with CW during a disruptive event: John is alerted by CW that his usual bus stop is inaccessible because of pluvial flood waters. CW further redirects him to an unaffected bus stop that services his route. John follows the global positioning system directions to the alternative bus stop and gives feedback about the quality of the service. The second scenario, illustrated in Figure 2(b), focuses on the urban management perspective in relation to adapting to the same disruptive event. Ann, a customer service representative in the city council, observes flooding damage to property trending on the CW interface. On inspecting the issues and the content given by the users, she issues the best practice advice to those in the areas affected by the flooding. She can see both actual water level data from urban sensors as well as historical water levels in the area. Users upload footage, photos and updates detailing the situation, all of which gives Ann a complete picture of the prevailing conditions in the city by filtering out all but the most trustworthy users.

The qualitative research highlighted that while long-term strategic planning and major emergency management at the upper levels of government (both local and national) were well developed; communication and interaction at ground level, particularly between the citizens and citizen-facing personnel, could be strengthened. Indeed, while the city flood defence mechanisms and procedures were excellent, much of the city personnel relied upon tacit knowledge and analogue forms of communication. Our fieldwork helped identify where this ‘hidden’ knowledge could be usefully supplemented with participatory sensing, building upon the opportunity to both capture the tacit skills and experience of the city service providers as well as to better integrate individual forms of information and communication. The CW is designed to assist existing systems by overcoming these shortcomings. It is worth noting, however, that while all stakeholders agreed that the availability of fine-grained real-time data as proposed by CW would significantly improve disruptive events handling, the user validation process highlighted that citizens would be very unlikely to instal an application targeting disruptive events, as it would not occur to them in normal circumstances, and

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

67

A. Manzoor et al.

when such events occur, people would not consider looking for an application, installing it and learning to use it.

The subsequent stakeholders workshop led to the definition of a use case around encouraging sustainable urban behaviours. Indeed, although the importance of reducing energy consumption and waste is widely recognised, there is a huge gap between theory and practice. Wasteful behaviours such as lights on at night in office buildings, buses idling or windows opened in heated buildings are common, much of which can be, and are, observed by citizens. Conversely, a number of green initiatives such as Urban Community Gardens go un-noticed but could have broader positive impact with more visibility.

with the location of the facilities, John can also see a short text description of the available facilities and photographs. Additionally, John can view the overall ‘green rating’ of his area, which is currently light green. He can see that the adjoining area is a vivid green. John filters all of the tags on the map and can clearly see that there are very few ‘Biodiversity’ tags in his area compared to the other areas in the neighbourhood. On his commute to work, John notices an illegal dumping incident in front of a derelict property and sees his opportunity to tag an issue with the mobile application. The application automatically geo-tags the location and time of day the picture was taken. He is automatically awarded five points for tagging the issue and subsequently received another 15 in the next hour with locals rating the tag on the application. Ann from the city council is notified of the incident by CW and dispatches workers to collect the garbage. Once this is done and reported in the application, the green rating of the area improves. Don, a leader of a community garden uses CW to motivate the garden community. Therefore, he occasionally posts several pictures of the community garden, showing off the work that was undertaken and how beautiful the garden is during bloom. Moreover, the sensors installed in the garden indicate the soil temperature and humidity in real-time. Thus, Don and his colleagues can organise the watering of the plants according to their actual needs. At harvest time, in early October, Don organises a Harvest Festival. He tags the event on the CW map. It receives many up-votes from community garden members and members of the wider community who support the event.

4.2. Approach

4.4. Outcomes

The design process investigated whether existing fixed sensors as well as the power of the crowd can be exploited to signal wasteful practices and highlight green initiatives, turning citizens into sensors to create a dynamic green urban map Dublin and eventually creating the impetus for greener cities. In particular, the stakeholder group was extended to include groups and individuals with particular interest in sustainability, such as community garden communities and green activists. Similarly to the other use case, the qualitative research investigated types of data that could be used, both from open data and participatory sensing, but also what other data might be useful, sustainability metrics most likely to engage users to participate, and how to enable sustained participation.

The quality research highlighted that there is a significant appetite for this use case from all stakeholders. Significant challenges were however identified around ensuring sustainable user engagement. To address this, a review of incentivisation mechanisms was undertaken, and a combined approach of rewards and gamification was chosen. In the proposed design, citizens are engaged in an urban game, where they compete in their daily sustainable behaviour. This is achieved by awarding points and incentives to users that highlight environmental issues and positive actions in the city.

4. GREENWATCH Following the output of the FloodWatch use case, further consultation with stakeholders, and in particular with citizens (ranging from teenagers to interested citizens, community and green activists), was arranged to investigate a use case that would encourage the ongoing use of CW. Such a use case can be used not only to validate our hypothesis of using city data to improve city resources management but also as a basis to embed the functionalities of a disruptive application use case, having already installed the CW platform and being familiar with its interface, users would be much more likely to use it in times of crisis. 4.1. Motivation

4.3. Personas and scenarios John has recently moved into a new apartment. He is interested in green initiatives in the area such as somewhere to grow some vegetables and to compost organic waste, as his apartment does not facilitate brown bin collection. CW highlights green resources such as local recycling, composting and community garden facilities in the city. Along 68

5. QUALITATIVE RESEARCH CONCLUSIONS This section discusses the overall output of the qualitative research in terms of the overall requirements on CW as well as detailed technical design challenges. 5.1. Overview The design approach adopted enabled us to identify design requirements for CW to be potentially adopted at the Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

city level, and we were able to subsequently design and deliver well-received citizen design workshops in followup research. Overall, our qualitative research provided us with the content, understanding and dynamic flexibility to analyse our data and design for a variety of following inter-related angles and informational hierarchies:  Operational hierarchy (chain of command)  Chronological hierarchy (idealized or actual flow of how data/decisions/actions occur; not always linear)  By data type/event management (historical, real-time, anticipatory or open, fixed/closed, inferred events)  Priority of major emergency management (e.g. life/limb, property/infrastructure, physical/emotional health, economic development or prevention, preparation, response, recovery) Overall, the strengths, weaknesses, opportunities and threats for CW were identified, as presented in Figure 3. In particular, CW has the potential to capture tacit knowledge, improve real-time two-way communication with citizens by creating a dynamic feedback channel, but stakeholders identified that such a system should integrate with existing services to realise its full potential. 5.2. Design challenges The CW design team anticipated different technical design challenges during the CW design and proof-of-concept implementation process. The CW stakeholder workshop participants also mentioned civic and social issues. Here, we briefly discuss those challenges and issues.

5.2.1. Urban-scale sensing. To gather real-time city data, along with fixed sensing, CW needs to take advantage of participatory sensing. Participatory sensing is a ubiquitous crowd-sourcing-based data collection technique that enables citizens, acting alone or in groups, to use their mobile devices to collect and transmit data. It is a powerful and cost effective tool to deliver city services with optimal use of resources [12]. However, because citizens perform many other tasks in their everyday life, they might not be motivated enough to actively participate in data collection campaigns. Data collection and transmission may also consume a lot of energy on a battery-constrained mobile device. To avoid these problems, CW is required to use different mechanisms to give incentives to citizens for effective data collection. These incentives can be citywide recognition of effective data contributors, wining virtual city titles, providing insight on the purpose of data collection and micropayments. Citizens may also have intrinsic motivations to contribute data, such as social contact and time passing, and extrinsic motivations, such as the satisfaction that they are helping to make their city a better place to live. CW needs to take advantage of these different aspects of human nature to conduct an effective data sensing campaign. 5.2.2. Data storage and dissemination. The CityWatch urban sensing system collects a new class of ‘geo-tagged’ high volume of data. In order to process this data and make it available to a new range of urban applications, an efficient data storage and dissemination system is of the utmost importance. To address this challenge, CW needs to implement the following features.

Figure 3. CityWatch strengths, weakness, opportunities and threats (SWOT) diagram.

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

69

A. Manzoor et al.

Scalability CW should provide the ability to integrate with other city data sets and city infrastructure such as city fixed sensor networks, city transport data feeds, traffic sensors and environmental sensing data feeds. Therefore, the application should be able to handle a continuously increasing volume of data. Capability of the data storage systems to store this data is crucial. The throughput and the speed of access to urban data must also be scalable. All these factors should scale on demand, growing with the number and volume of supported applications, whilst causing the minimum possible increase in the system’s complexity or overhead. Cost-effective tiered storage When it comes to storing and retrieving the expected massive amounts of data in a future city application that can be very temporal in its nature, the first question developers should ask themselves is ‘How long do I need wait to get the data I need?’ The question that comes soon after is ‘What is the cost of storing the data?’ To answer these questions, CW design should implement a cost-effective efficient data storage system that meets the needs of different data users. Self management The urban data should be usable across several city use cases, such as flood response activities or the efficient use of city resources. To fulfil this requirement, the CW data storage system should therefore be smart enough to automatically handle vast data streams while simultaneously taking into account users priorities, some of which might conflict with other users priorities. Adaptive context data dissemination and provisioning where data is moved around the system differently depending on the contextual requirement of the system is one possible avenue to address this challenge. High availability and wide accessibility High availability and short access latency are required to ensure that city stakeholders can rely on the platform. 5.2.3. Trust, security and privacy. The characteristics of urban-scale sensing environments make CW vulnerable to the collection of low-quality data. Collected data may be redundant, as many sensors may collect information about the same event in the city; inaccurate, as the sensors installed at public places in the city may start to malfunction because of environmental conditions or malicious handling; and conflicting, as different citizens contributing data may perceive an event differently. Sensors contributing high-quality data may not be available indefinitely, such as mobile sensors installed on smart phones, or smart vehicles might move away from the data collection unit or human sensing participants might opt out for convenience reason or due to a lack of motivation. Low-quality data may generate false events in the city, leading to severe effects on the functionality of CW, such as city management authorities inaccurately perceiving the current situation in the city and incorrectly allocating resources. These circumstances mean that CW should be 70

designed to be able to identify high-quality sensing streams and dynamically find out alternative data sources [13]. In this regard, CW faces a challenge to identify and validate high-quality sensing streams by evaluating their trustworthiness to contribute high-quality data that truly predicts the current situation in the environment. During urban-scale sensing, human sensing participants carrying a smart phone embedded with a multitude of sensors may also share information about their current location, for example, home or office; work routines, for example, the time at which they arrive at and leave their work place; daily activities, for example, a visit to their doctor; and social life, for example, who they meet and talk to. They may also share their image and voice (through noise detection for example). Sharing this information raises concerns about their privacy, compromising their freedom of speech, association and fundamental human rights [14, 15]. The CW design needs to pay special attention to maintaining participants’ privacy by informing them of the intended use of the information that they share; enabling them to control the nature, granularity and amount of this information, and empowering them to sanction when, where and with whom they will share it. Revealing private information may also put the safety and security of sensing participants from thieves and stalkers. CW requires a mechanism to maintain the integrity and security of the sensed data, so that it is only available to authorised applications and data consumers. Maintaining the privacy and security of sensing participants and making them aware of the intended use of their shared information may motivate the sensing participants to actively participate in the sensing tasks.

6. CITYWATCH: SYSTEM OVERVIEW This section presents an overview of the prototype CW architecture and some of its salient features addressing some of the design challenges presented in the previous paragraphs (other challenges have yet to be addressed, as discussed in Section 10). Figure 4 shows the CW system architecture. The CW middleware and the CW application server are the two main components of the CW framework. The CW middleware works as an intermediary layer between the data producers and the data consumers and provides an interface for easy and scalable data collection and dissemination. The CW application server manages the CW Web † and smart-phone-based applications. The Web applications allow only to view data, whereas the smart phone versions can be used both to view and to share data. The CW application server adds a layer of abstraction between applications and the middleware. In the remainder of this section, we present these components in more details.



see, e.g. http://www.citywatch.ie/

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

Figure 4. CityWatch architecture.

Figure 5. CityWatch middleware components.

6.1. Middleware The CW middleware is a layer of interoperable and scalable software components that enables smart city applications to discover and access urban sensor resources. Figure 5 shows the details of these components. The CW

middleware interface for smart city applications or data consumers allows them to interact with sensor resources in a well-defined, systematic way. Similarly, the CW middleware interface for data producers or sensor resources enables them to share data resources. Figure 6 shows the CW middleware interface for data producers and data consumers. In the subsequent paragraphs, we explain the CW middleware components and the CW middleware interfaces in detail. The Sensor Manager is responsible for managing data producers in the CW middleware. Data producers include the urban entities that produce data, such as fixed sensors and mobile sensors. The CW middleware allows data producers to share urban data through a well-defined interface as shown in Figure 6. A data producer that wants to connect to the CW middleware to share data will call RegisterDataStream to let the middleware know its intention to share data. It also sends information about the data types it can provide, their representation format, the frequency of updates and the measurement unit, as well as a

Figure 6. CityWatch middleware interface.

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

71

A. Manzoor et al.

Figure 7. CityWatch application server components.

call back function. The CW middleware calls this function to pull data from the data producer when required. Data producers can also use the AddData and AddDataList interfaces to push data to the middleware. If a data producer decides to stop sharing data, it can call RemoveDataStream to unregister the data stream. The Subscription Manager keeps track of the consumers’ data subscriptions and sends data to them. Data consumers include the entities that consume data to perform different functionalities. These entities can be smart city applications or other data services. The CW middleware provides an interface for data consumers to obtain information about available data in the middleware and set subscriptions for that data. Figure 6 shows the details of that interface. Data consumers can call GetDataType to get information about the available data types in the middleware. Data consumers can further call DescribeDataType to query metadata, such as data models. Finally, data consumers can use the SetSubscription interface to set data subscriptions by specifying the requested data type, update frequency and a call back function to receive data. Data consumers can also call GetData to receive data. Once data are collected from different sensing systems, the Trust Evaluator quantifies the trustworthiness of the data producers as the belief in their capabilities to collect high-quality data as described by Manzoor et al. [16]. The Data Aggregator detects and removes duplicate and conflicting information. The CW middleware supports pushbased and pull-based data communication techniques. Data producers can push data to middleware or set a call back function to allow the middleware to pull data when needed. Similarly, data consumers can also use the CW middleware interfaces to pull data or set subscriptions to allow the middleware to push data. The Data Communication module is responsible for providing this functionality. The Data Query and Storage module is responsible for storing and querying data in the database. 6.2. Application server The CW middleware provides the data collection and dissemination functionality for multiple applications. The CW application server provides additional functionality on top of the CW middleware to CW applications. This functionality includes the user management, report handling, 72

notification management, gamification, date query and storage, reputation tracking, privacy and security. Figure 7 illustrates the main components of the CW application server. The Report Manager contains the necessary methods for storing, processing and retrieving user-submitted reports. The Privacy and Security module provides secrecy to the exchanged data, along with image anonymisation by both removing faces and stripping image metadata. The User Management component provides general management of the user data, groups and areas. The Gamification component provides the main functions to construct and manage a game that incentivises users to use the application more. There is also the Data query and Storage component, which manages the procedures for storing, retrieving and processing information. Moreover, the Reputation manager rates the credibility of each user’s feedback. Finally, the Notification component provides location-aware user notifications within specific time frames. The data that are submitted to CW is geo-tagged, which clearly discloses the identity of the user that submitted them. Because the geolocation of users is sensitive, as it may disclose many aspects of user’s life, CW tries to share as little information as possible. In this context, CW does not disclose the source of sensor measurements to anonymise them. Furthermore because the users can upload reports, which are disclosing their location at a specific time, CW will provide the name of a user only to users of the same group, masking all the others. One more precautionary measure in CW for providing more privacy is the anonymisation of images. The reports that are sent to CW usually contain images. However, when taking a picture on the street, one can accidentally capture other people in the frame. Because this action was made without their consent, CW attempts to detect faces in the submitted pictures with some standards methods. If a face is detected, then a black box is drawn on top of it, thus masking whoever is depicted. Additionally, photos from mobile phones very often contain metadata providing information such as, the geolocation, camera id, lens id and phone model. Although the geolocation might be disclosed with the report, the other metadata can be used in correlation attacks, allowing attackers to deduce further information. Therefore, each photo submitted to CW is stripped from any metadata. For security, CW is using Secure Sockets Layer (SSL) connections, so user-submitted information is encrypted with standardised and secure algorithms. Additionally, OAuth‡ is used for registering users through their Google accounts. This provides some guarantees that the users are real and not bogus and, additionally, allows the users to see the minimal information that is requested from their Google accounts from CW. For further details about these components, interested users may refer to [9].



www.oauth.net

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

Figure 8. Implementation diagram.

7. IMPLEMENTATION A general overview of the implementation is illustrated in Figure 8. We used Java EE to implement the CW middleware functionality described in Section 6.1 and deployed it on the Google App Engine (GAE). The Java API for XMLbased Web services Java specification request 224 is used to implement Simple Object Access Protocol-based Web services providing the middleware interface to data producers and data consumers as described in Figure 6. Web Service Definition Language files describing the CW middleware interface details and schema files describing the data types used in Web Service Definition Language file can be accessed at the GAE. Data consumers and data producers share data with the middleware in XML format. A common XML schema describing the data representations is shared among different components. Data producers generate data objects according to the XML schema. Once the middleware receives an XML data object, it generates a unique identifier for that measurement and adds it to the data object. We use the Java API for XML-based remote procedure call Java specification request 101 for data validation and XML-Java object binding. Subsequently, the data are stored in a Google Structured Query Language (SQL) database. We use the Hibernate and JAVA Persistence API 2.0 for object relation mapping and interaction with the SQL database. The CW application server sits on top of the CW Middleware, providing a RESTful API to mobile applications. It is implemented in Python 2.7 and deployed on the GAE as well. For storing user information, the application server is using a Google SQL database, while images, which have a significantly bigger size, are uploaded to the Google Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

Cloud Storage. Because these two services provide automatic scaling and very good performance, handling the data and load balancing are forwarded to Google’s infrastructure, which scales them smoothly and transparently. Generally, all input calls and output data are XMLformatted, to provide interoperability and easy parsing, independently of the used client. All the calls are made through Hypertext Transfer Protocol Secure to provide secrecy of the exchanged data with standard and secure algorithms. Additionally, only calls from users that have submitted their credentials through OAuth 2.0 are permitted. Because the vast majority of Android users has a Google account, this account and its credentials are used to authenticate the user to the application server. As discussed, to provide some anonymity to the reports, each uploaded image is processed using OpenCV§ to remove detected faces. Afterwards, any metadata is removed from the photos and they are uploaded to the Google Cloud Storage. To minimise bandwidth costs, the application server returns only links to images in Google Cloud Storage. This way, the exchanged messages are short, and the application programmer chooses when pictures should be received and displayed. To display the functionality that our infrastructure is providing, a mobile application has been developed using HTML5 and JavaScript, using the PhoneGap¶ framework. The application has been packaged for Android OS, but as it is written in HTML5, it can be easily ported to any other widely used mobile OS, such as iOS, Blackberry, or Windows Phone. The choice of Android to pilot our application § ¶

http://opencv.org/ www.phonegap.com

73

A. Manzoor et al.

(a) Current location

(b) User group

(c) User profile

Figure 9. Screen shots of CityWatch mobile application: (a) current location, (b) user group and (c) user profile.

was motivated by the fact that it currently has the biggest market share|| . The interface is quite intuitive and allows users to easily report positive or negative events and access city sensors as evident from some of screenshots shown in Figure 9. The mobile application is authenticating users using their credentials from their Google account through OAuth 2.0 and forwards all their requests encrypted to the application server, using a Hypertext Transfer Protocol Secure connection and the provided RESTful API. The application requires minimal access to the users’ accounts, more precisely, it demands only access to their name and email. Moreover, the application is using a user’s location only when he submits a report, thus keeping in line with the principle of data minimisation.

8. USER TRIAL AND RESULTS In this section, we report the preliminary results of our ongoing user trial of a GreenWatch application built on top of our framework. We organised two workshops in Dublin City Council and Trinity College Dublin to invite volunteers to participate in our trial. Currently, 55 users have volunteered to take part in our trials. They have installed the application on their own phones and are collecting and reporting the data during their everyday life. Considering the civic importance of urban environment, stakeholders interest and wide availability of related data in everyday life, we asked our volunteers to report urban practices and events affecting the environment sustainability positively and negatively. They are also using the application to view sensor data. Figure 10 shows the distribution of the number of CW users’ interaction with the application and subsequently the ||

http://www.idc.com/getdoc.jsp?containerId=prUS24108913

74

back-end servers against the number of users performing those interactions. The CW users’ interactions are counted as the total number of their reports capturing positive and negative urban practices, their actions to resolve reported problems, votes up or down of reports, sharing of reports on social media, queries on sensors data and queries of the green score of an area. The graph shows that a few users have a very high number of interactions with the application, such as one user who has interacted with the application 48 times, whereas a large number of users rarely interact with the application. We can also find the same pattern in the graphs shown in Figures 11 and 12. Figure 11 shows the cumulative distribution of the total number of reports of negative and positive urban behaviours. We can observe that about 14 sensing participants contributed at least one report. The number of sensing participants decreases as the number of sensing reports increases, and we can find that very few sensing participants contributed over 20 reports. Figure 12 shows the distribution of the issues by CW users’ in the different issue categories. (Note, the tag categories are pre-defined in the application, and users can only send information in one of these categories). We can observe that users took different level of interest in tagging those issues. CW users showed more interest in sharing reports about Biodiversity and Litter issues in the city than reports about the availability or lack thereof of cycling resources or the pedestrian accessibility of different areas of the city. The very high number of reports shared about the presence of Biodiversity and Litter in the city and the very low numbers of reports contributed about green events in the city might be explained by two reasons. Firstly, Biodiversity and Litter are the top options to share positive and negative reports (respectively) in the CW application user interface, whereas NoPedestrianAccess and EventDiscovered lie at the bottom of the CW user interface. Secondly, Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

Figure 10. User participation cumulative distribution in terms of number of interactions.

Figure 11. User participation cumulative distribution in terms of number of reports.

sensing participants may have different level of interests in sharing data about different topics. Considering these observations, it is highly recommended that special consideration should be paid to the user interface of participatory sensing based applications and interests of a particular community to effectively collect urban data. We can observe that the graphs plotted in Figures 11– 13 follow the power law. Power-law phenomena are also evident in social interactions and social networks. Observing the presence of a power-law-based distribution in the collected data, we can establish that all the sensing participants are not sharing data at the same rate. Few participants have a high interaction with our participatory-sensingbased application and are sharing a lot of information about different aspects of urban life. Whereas a large number of people rarely interact with the application and share any information. Sitting between these two extremes, a number Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

of people interact occasionally with the application. Active users can motivate the moderately active user group to increase their participation and engage the less active users. Note that a power law pattern is also present in Figure 12. Figures 14 and 15 show data contributed by the five most active CW users. Figure 14 shows the proportion of positive and negative reports, whereas Figure 15 shows the proportion of detailed tags. We can observe that different users contributed data in different patterns. We can observe in Figure 14 that User 3 mostly contributed negative events reports, whereas User 4 only contributed positive reports. However, if we look at the total number of reports contributed by all users, we can find that positive and negative reports have been shared in equal proportion. We find the same pattern in Figure 15. Different users have also shared reports about different issues with different intensity, though overall most of the issues have been 75

A. Manzoor et al.

Figure 12. User participation for different issue categories.

Figure 13. Users’ queries to view data from different sensors.

shared with a similar proportion. This observation hints that although different users may share data about different issues according to their own interests in a particular issue, if there is a high number of users sharing data with an application, overall, all the issues might be shared in similar proportions.

9. STATE OF THE ART With a persistently increasing rate of urbanisation, it is expected that 6.3 billion people will be living in cities, towns and conurbations in 2050 [17]. This unprecedented increase in urban population will pose new challenges to provide urban services in the domain of energy, infrastructure, mobility and economy to provide a sustainable and liveable place for citizens [18, 19]. Consequently, sev76

eral governmental, industrial and academic collaborations have been initiated to address these challenges and have started exploiting advancements in information and communication technology to improve the management of the limited city resources [20, 21]. This section provides an overview of applications developed to take advantage of ubiquitously available smart phones to collect data and provide useful urban services. The availability of a number of sensing systems on smart phones have made them an important tool to collect data from social and urban environments [22]. Recently, many applications have been developed to take advantage of these ubiquitously available sensing systems to collect data and provide useful services in the domains of urban life [23, 24], environment sustainability [25, 26], air quality and noise monitoring [27, 28] and health care [29, 30]. These applications, however, Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

Figure 14. Proportion of positive and negative issues tagged by the top five CityWatch users.

Figure 15. Proportion of different issues tagged by the top five CityWatch users.

directly connect to the smart phone sensors to collect data for their own use [31], which is very time-consuming and resource-intensive. It is a challenging task for application developers to deal with different kind of communication protocols and sensing systems [32]. They may also have to learn different languages to develop software for different platforms. Consequently, a framework to perform those tasks for the application developers is expected to be highly invaluable [33]. Few research efforts have targeted the provision of a generic data collection and dissemination framework for rapid application development. Here, we present the comparative analysis of those research efforts with our proposed system. Trossen and Pavel presented a platform, Nokia Remote Sensing, which allows mobile devices to be the part of sensing frameworks and works as gateways to forward sensor data [34]. Nokia Remote Sensing is based on Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

the publish–subscribe communication paradigm. However, unlike in our work, the details of the interface to enable sensors to send data and applications to query data from the middleware are not provided. Furthermore, only fixed sensors, temperature, humidity, pressure and dew point are used in their test bed implementation of their platform, whereas, along with fixed sensing, our platform also enables users to contribute data. Reddy et al. proposed a network service architecture for participatory sensing [35]. They emphasise that participatory sensing participants selection, incentivisation and tasking for effective data collections are the main challenges of a participatory sensing framework. They stress that a sensing framework should pay special attention to the evaluation of the trustworthiness of the data sources and the protection of the privacy of sensing participants. However, they have not implemented their proposal. 77

A. Manzoor et al.

Tuncay et al. [36] presented a distributed framework to select sensing participants and to collect data. They used profile-cast [37], a behaviour-oriented protocol that aims to send messages to the nodes matching a certain target profile, to select sensing participants suitable for a given data collection campaign. They propose to use data sinks in a particular area to collect data. Subsequently, the data are conveyed to the organisers of the sensing campaign. However, their framework targets a direct sensing campaign for each data consumer. They did not make provision for the collected data to be distributed to multiple applications in their framework. Sheng et al. [38] targeted the provision of sensing services using mobile phones via a cloud computing system. They aimed to support different mobile-phone-based sensing applications through their service. They also discussed different design choices for such service and emphasised that energy efficiency and an effective incentivisation mechanism for sensing participants are the main requirements of such a service. They have also discussed their hypothetical models to achieve those capabilities. However, they have not presented any information about the design and implementation of their system. Joki et al. [39] presented a framework for participatory data collection on smart phones—Campaignr. The framework enables individuals and groups of people interested in collecting urban data to initiate a data collection process. However, the framework was specifically developed for Symbian S60 3rd edition phones. Consequently, only the owners of that particular smart phone can take advantage of their framework. The framework did not allow a generic data consumer or smart phone application to collect data according to their own requirements. Rather than proposing a complete data collection and dissemination framework, some research efforts have also targeted specific issues in participatory sensing. A special emphasis has been put on preserving the privacy of sensing participants [40, 41] and evaluating their trustworthiness [16, 42]. Other work has also been undertaken to characterise the sensing systems and to ensure the collection of high quality data [43]. Although these solutions can be used to improve the performance of a generic data collection and distribution framework, they are not directly related to our work. Existing research literature lacks a generic data collection and dissemination system.

10. CONCLUSION This article presented the qualitative research techniques used in the design of CW, an urban scale data dissemination and sensing framework. It highlighted how stakeholders and domain experts interviews, field observations and a user-led design workshop were used to co-design solutions to specific city challenges and, eventually, to derive requirements for the overall framework. We also presented a prototype framework architecture and implementation, as well as giving an overview of one of the applications developed using the framework, and results of an ongoing trial. 78

We showed that different users have shared data about different issues with a different intensity. We find that overall user participation follows the power law, and that although there is a group of very active users, a large number of users rarely interact with the application. Active users can play a role in motivating the moderately active user group to increase their participation and engage the less active users. Although these results are very encouraging, we have yet to address a number of the challenges that we have identified, by extending the capabilities of our framework to include more features, such as opportunistic task assignment by dynamically finding out the most suitable group of sensing participants to gather information about a specific issue, sensor stream quality validation and improved privacy and security.

ACKNOWLEDGEMENT This work was supported by the Science Foundation Ireland under the Principal Investigator research programme 10/IN.1/I2980 ‘Self-organising Architectures for Autonomic Management of Smart Cities’.

REFERENCES 1. World urbanization prospects. the 2005 revision, pop. division, department of economic and social affairs, un. 2. Dirks S, Keeling M. A vision of smarter cities: how cities can lead the way into a prosperous and sustainable future. IBM Institute for Business Value, June 2009. pp. 1–6. 3. Catarci T, de Leoni M, Marrella A, Mecella M, Salvatore B, Vetere G, Dustdar S, Juszczyk L, Manzoor A, Truong H-L. Pervasive software environments for supporting disaster responses. IEEE Internet Computing 2008; 12(1): 26–37. 4. Perera C, Zaslavsky A, Christen P, Georgakopoulos D. Sensing as a service model for smart cities supported by internet of things. Transactions on Emerging Telecommunications Technologies 2014; 25(1): 81–93. 5. Cimmino A, Pecorella T, Fantacci R, Granelli F, Rahman TF, Sacchi C, Carlini C, Harsh P. The role of small cell technology in future smart city applications. Transactions on Emerging Telecommunications Technologies 2014; 25(1): 11–20. 6. Cook DJ, Das SK. Pervasive computing at scale: transforming the state of the art. Pervasive and Mobile Computing 2012; 8(1): 22–35. 7. Lane ND, Miluzzo E, Lu H, Peebles D, Choudhury T, Campbell AT. A survey of mobile phone sensing. IEEE Communications Magazine 2010; 48(9): 140–150. 8. Carayannis EG, Campbell DFJ. ’Mode 3’and’Quadruple Helix’: toward a 21st century fractal innovation Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

A. Manzoor et al.

9.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

20. 21.

22.

ecosystem. International Journal of Technology Management 2009; 46(3): 201–234. Manzoor A, Patsakis C, Bouroche M, McCarthy J, Mullarkey G, Clarke S, Cahill V. Data sensing and dissemination framework for smart cities. In MobilWare2013, IEEE: Bologna, Itlay, 2013. Cooper A, Reimann R, Cronin D. About Face 3: The Essentials of Interaction Design. John Wiley & Sons: Indianapolis, Indiana, 2012. Saffer D. Designing for Interaction: Creating Smart Applications and Clever Devices. New Riders: Berkeley, California, 2010. Burke JA, Estrin D, Hansen M, Parker A, Ramanathan N, Reddy S, Srivastava MB. Participatory sensing. Technical Report, Center for Embedded Network Sensing, UCLA, 2006. Available from: http://escholarship.org/uc/ item/19h777qd [25 November 2013]. Reddy S, Estrin D, Srivastava M. Recruitment framework for participatory sensing data collections. In Pervasive Computing. Springer, 2010; 138–155. Shilton K. Four billion little brothers?: privacy, mobile phones, and ubiquitous data collection. Communications of the ACM 2009; 52(11): 48–53. Rebollo-Monedero D, Bartoli A, Hernández-Serrano J, Forné J, Soriano M. Reconciling privacy and efficient utility management in smart cities. Available from: http://www.cttc.es/resources/doc/121210-upc-ieeecommun-mag-120831a-7088025322897495396.pdf [25 November 2013]. Manzoor A, Asplund M, Bouroche M, Clarke S, Cahill V. Trust evaluation for participatory sensing. In Mobile and Ubiquitous Systems: Computing, Networking, and Services. Springer: Berlin, Heidelberg, 2013; 176–187. Un. world urbanization prospects: the 2011 revision., 2011. Available from: http://esa.un.org/unpd/wup/index. htm, [August 2013]. Shapiro JM. Smart cities: quality of life, productivity, and the growth effects of human capital. The Review of Economics and Statistics 2006; 88(2): 324–335. Toppeta D. The smart city vision: how innovation and ict can build smart, livable, sustainable cities, The Innovation Knowledge Foundation, 2010. Available from http://www.thinkinnovation.org/file/research/23/en/Top peta_Report_005_2010. pdf [25 November 2013]. City science, mit media lab. Available from: http://cities. media.mit.edu/about/cities [03 June 2013]. Smart city lab, The slexandra institute: Denmark. Available from: http://www.alexandra.dk/uk/labs/ smart-city-lab/pages/smart-city-lab.aspx [03 June 2013]. Miluzzo E. Smartphone sensing. Ph.D. Thesis, DARTMOUTH COLLEGE Hanover, New Hampshire, 2011.

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett

23. Horanont T, Shibasaki R. An implementation of mobile sensing for large-scale urban monitoring. In Proceedings of Urbansense08, Raleigh, North Carolina, USA, November 4 2008; 51–55. 24. Froehlich J, Neumann J, Oliver N. Measuring the pulse of the city through shared bicycle programs. In Proceedings of UrbanSense08, Raleigh, North Carolina, USA, November 4 2008; 16–20. 25. Mun M, Reddy S, Shilton K, Yau N, Burke J, Estrin D, Hansen M, Howard E, West R, Boda P. Peir, the personal environmental impact report, as a platform for participatory sensing systems research. In Proceedings of the 7th International Conference on Mobile Systems, Applications, and Services, MobiSys ’09, Krakow, Poland, 2009; 55–68. 26. Rana RK, Chou CT, Kanhere SS, Bulusu N, Hu W. Earphone: an end-to-end participatory urban noise mapping system. In Proceedings of the 9th ACM/IEEE International Conference on Information Processing in Sensor Networks. ACM: New York, NY, USA, 2010; 105–116. 27. Mendez D, Perez AJ, Labrador MA, Marron JJ. P-sense: a participatory sensing system for air pollution monitoring and control. In 2011 IEEE International Conference on Pervasive computing and communications workshops (PERCOM Workshops), IEEE: Seattle, USA, 2011; 344– 347. 28. Maisonneuve N, Stevens M, Niessen ME, Steels L. Noisetube: measuring and mapping noise pollution with mobile phones. In Information Technologies in Environmental Engineering. Springer: Berlin, Heidelberg, 2009; 215–228. 29. Reddy S, Parker A, Hyman J, Burke J, Estrin D, Hansen M. Image browsing, processing, and clustering for participatory sensing: lessons from a dietsense prototype. In Proceedings of the 4th Workshop on Embedded Networked Sensors. ACM: New York, NY, USA, 2007; 13–17. 30. Boulos, Wheeler S, Tavares C, Jones R. How smartphones are changing the face of mobile and participatory healthcare: an overview, with example from ecaalyx. Biomedical Engineering Online 2011; 10(1): 24. 31. Dutta P, Aoki PM, Kumar N, Mainwaring A, Myers C, Willett W, Woodruff A. Common sense: participatory urban sensing using a network of handheld air quality monitors. In Proceedings of the 7th ACM Conference on Embedded Networked Sensor Systems, SenSys ’09, Berkeley, California, USA, 2009; 349–350. 32. Martin S, Diaz G, Plaza I, Ruiz E, Castro M, Peire J. State of the art of frameworks and middleware for facilitating mobile and ubiquitous learning development. Journal of Systems and Software 2011; 84 (11): 1883–1891. 79

A. Manzoor et al.

33. Payton J, Julien C. Integrating participatory sensing in application development practices. In Proceedings of the FSE/SDP Workshop on Future of Software Engineering Research. ACM: New York, NY, USA, 2010; 277–282. 34. Trossen D, Pavel D. NORS: an open source platform to facilitate participatory sensing with mobile phones. In Fourth Annual International Conference on Mobile and Ubiquitous Systems: Networking Services, 2007. MobiQuitous 2007, Philadelphia, USA, 2007; 1–8. 35. Reddy S, Samanta V, Burke J, Estrin D, Hansen M, Srivastava M. Mobisense—mobile network services for coordinated participatory sensing. In International Symposium on Autonomous Decentralized Systems, 2009. ISADS ’09, Athens, Greece, 2009; 1–6. 36. Tuncay GS, Benincasa G, Helmy A. Autonomous and distributed recruitment and data collection framework for opportunistic sensing. ACM SIGMOBILE Mobile Computing and Communications Review February 2013; 16(4): 50–53. 37. Lane ND, Eisenman SB, Musolesi M, Miluzzo E, Campbell AT. Urban sensing systems: opportunistic or participatory?. In Proceedings of the 9th Workshop on Mobile Computing Systems and Applications, HotMobile ’08. ACM: New York, NY, USA, 2008; 11–16. [Online]. Available from: http://doi.acm.org/10.1145/ 1411759.1411763.

80

38. Sheng X, Xiao X, Tang J, Xue G. Sensing as a service: a cloud computing system for mobile phone sensing. In Sensors, 2012 IEEE, IEEE: Taipei, 2012; 1–4. 39. Joki A, Burke JA, Estrin D. Campaignr: a framework for participatory data collection on mobile phones. Technical Report, Center for Embedded Network Sensing, UCLA, 2007. Available from: http://escholarship.org/uc/ item/8v01m8wj [25 November 2013]. 40. Huang KL, Kanhere SS, Hu W. Preserving privacy in participatory sensing systems. Computer Communications 2010; 33(11): 1266–1280. 41. De Cristofaro E, Soriente C. Short paper: PEPSI— privacy-enhanced participatory sensing infrastructure. In Proceedings of the Fourth ACM Conference on Wireless Network Security, ACM. Hamburg, Germany, June 15–17, 2011; 23–28. 42. Dua A, Bulusu N, Feng W-C, Hu W. Towards trustworthy participatory sensing. In HotSec09: Proceedings of the Usenix Workshop on Hot Topics in Security, Montreal, Canda, August 11 2009; 8. 43. Manzoor A, Truong H-L, Dustdar S. Quality of context: models and applications for context-aware systems in pervasive environments. The Knowledge Engineering Review, Special Issue on Web and Mobile Information Services 2014, Article is available online https://svn.compas-ict.eu/staff/truong/publications/2010/ mtd-ker2010.pdf [25 November 2013].

Trans. Emerging Tel. Tech. 25:64–80 (2014) © 2014 John Wiley & Sons, Ltd. DOI: 10.1002/ett