Context-Aware Advertising Systems for Mobile Devices

James Wooltorton

Abstract

Mobile devices, such as PDAs and mobile phones, present excellent opportunities for directly targeting advertising at end-users. By utilising context through context-aware system it should be possible to produce highly personalised advertising and avoid bombarding users with large amounts of unsuitable, badly matched data.

This paper will discuss the areas associated with the development of a context-aware system for pushing product advertisements to mobile devices using contextual information to determine the most appropriate product from a retailer’s catalogue. This review will discuss ubiquitous computing, context-awareness, contextual data, location awareness, mobile advertising systems, as well as looking at possible approaches and reviewing some existing applications.

Introduction

Mobile devices are by their very nature private and personal devices. How many of us have received a text message only to be disappointed by annoying service messages or unsolicited advertising? Context-aware systems present excellent opportunities for directing highly targetable advertisements at end-users through these devices. This can be achieved by utilising contextual information from these devices and by learning end-user preferences and habits. Although, it is critical for the acceptance of any such system that is does not feel intrusive, whilst also putting considerations for security and privacy foremost. To address these problems this paper focuses on the major challenges associated with developing context-aware systems for mobile advertising.

We live in a digital age, surrounded by low-cost, ever-increasingly powerful computing devices, which are becoming progressively more integrated into our everyday lives. These embedded devices are already in our cars, homes, workplaces, refrigerators, and with the current technological trends it seems inevitable that they will commonly become both wireless and mobile. Such a scenario of large numbers of embedded and mobile devices, all invisibly communicating over an interconnected network, is an ideal environment for ubiquitous computing.

Ubiquitous computing

Ubiquitous computing names the third wave in computing (figure 1), just now beginning. First were mainframes, each shared by lots of people. Now we are in the personal computing era, person and machine staring uneasily at each other across the desktop. Next comes ubiquitous computing, or the age of calm technology, when technology recedes into the background of our lives.(Weiser, 1994)

Wesier first introduced ubiquitous computing as “the method of enhancing computer use by making many computers available throughout the physical environment, but making them effectively invisible to the user” (Weiser, 1993). Abowd et al extended this idea by saying “ubiquitous computing is the attempt to break away from the traditional desktop interaction paradigm and move computational power into the network and environment that surrounds the user”. (Abowd et al., 1998). Both Wesier and Abowd saw the potential in this radical new approach to computing, heralding the idea of harnessing the power of these devices, invisibly integrating them in order to enhance our modern lives. Weiser, perfectly sums it up by saying “It is invisible, everywhere, computing that does not live on a personal device of any sort, but is in the woodwork everywhere” (Weiser, 1994)

Figure 1: The Three Ages of Computing (Weiser, 1996)

There are many challenges which face developing truly ubiquitous systems, it will require “much new work in operating systems, user interfaces, networks, wireless, displays and many other related areas.” (Weiser, 1994) Central to ubiquitous computing are the key concepts of concurrency and parallelism, so from a developmental point of view existing methodologies are not suitable for ubiquitous systems, a view backed up by Borah, who highlights that “existing middleware for client server computing such as .NET, J2EE and CoBrA were not designed for ubiquitous computing”. (Borah, 2005) A developmental platform is required where these concepts are intrinsic and not bolted on to an existing solution.

Ubiquitous Environments

One such approach is the Indus platform (Borah, 2005), first conceptualised back in 2002, it is an attempt to embrace object-orientation principles and extend them to ubiquitous systems. Indus is a software agent platform, designed specifically for ubiquitous computing, it comprises of a programming language, libraries and containers or run time environments for cross platform distribution. Indus targets specific ubiquitous application groups; these include Intelligent Transport Systems, Supply Chain Management, Personal Area Networking and Control & Automation.

The Indus programming language extends the Java platform (Java API, 2002), introducing new concepts for context, and ports, as well as replacing the Java Interface type with two new interface types called Agent and Component, therefore every Indus class is either an Agent or a Component.

· An agent in Indus represents a role, additionally an agent can have several methods. Each method represents a task that can be implemented by an actor representing that role.

· Components are re-usable program building blocks, that implement static functionality of a system and compose other components.

In Indus, we can model any system as a set of concurrently executing agents. The Indus libraries include general purpose services for self configuration, service discovery, security, intelligent routing, transaction and policy management. Additionally, Indus compiles directly to native code on a number of systems, making it highly deployable. On enterprise platforms Indus compiles to Java byte code and runs using the JVM, whilst on mobile platforms it compiles directly to C code and binary executables.

Context and Context Awareness

Context

”The use of context is increasingly important in the fields of mobile and ubiquitous computing, where the user’s context is rapidly changing.” (Abowd et al., 1999)So what exactly is context? We can think of context as the circumstances under which a device is being used, and its situational information during events, or as summarised by Abowd et al, “ any information that can be used to characterise the situation of an entity, where an entity can be a person, place or physical or computational object.” (Abowd & Dey, 2002)

Korkea-aho provides some useful examples (Korkea-aho, 2000) of contextual data that can be utilised in a context-aware system, these include:

· spatial information - e.g. location, orientation, speed and acceleration

· temporal information - e.g. time of day, date, day of week

· environmental information e.g. temperature, weather conditions

· activity - e .g. shopping, dining, walking, working.

· social situations - e.g. people you are with or who is nearby.

· personal information - e.g. mood, blood pressure.

Context-Awareness

Abowd and Dey define context-awareness or context-aware computing as “the use of context to provide task-relevant information and/or services to a user.” (Abowd & Dey, 2002) Schilit says “context-aware computing devices react to their changing environment in an intelligent way so as to enhance the computing environment of the user.” (Schillit et al, 1994) It is the goal of any context-aware system to be able to interpret low-level contextual data and make assumptions about the user’s current situation.

A system can be thought of as context-aware if it can extract, interpret and use contextual information and adapt its functionality according to the current context of use (Pascoe 1997). Abowd et al (2000) and Troung et al (2001) identified the following questions, known as the five W’s, which any context-aware system needs to address: Who is using the system? What is he/she doing? Where is he/she located? When does the interaction take place? Why is the user doing that specific action?

Pascoe identifies that “the main challenge for any context-aware system lies in the complexity of capturing, representing and processing this contextual data.” (Pascoe, 1997) Although this looks at context-aware systems primarily from a technical stance, a view supported by (Kaasinen, 2002). In fact many context-aware projects cite more human factors, like privacy concerns (Hong 2004)(Kourouthanasis, 2002)(Newman et al., 1991) or simply the rejection of a system that feels intrusive (Want et al., 1995), as the key concerns for the success of a CA system.

Range Values

Another issue often associated with contextual data is dealing with range values, this is primarily because “context-aware queries rarely revolve around exact values” (Picco et al., 2005). Picco et al go further to identify that in context-aware systems queries are often “formulated over value ranges, and there is a need to deal with imprecise information coming from many sources”. For example, a system could represent an individual as aged 18 to 30 instead of their exact ages. Conventional programming languages only provide facilities for exact matching, a more adaptive approach is required in order to support data aggregation.

LighTS (Picco et al., 2005), is a lightweight Java API, which uses a tuple space model inspired by Linda (Gelernter, 1985), to provide a solution to this problem for context-aware applications. It provides emphasis on supporting range values and uncertainty in context-aware applications, by providing methods for specifying upper and lower bounds of range values for matching and using fuzzy logic techniques for dealing with uncertainty in data.

Context-Aware Applications

Context-aware systems offer entirely new opportunities for application developers and for end-users by gathering context data and adapting systems behaviour accordingly (Baldauf et al, 2005) Some example of early context-aware systems included the Active Badge (Want et al., 1992), Xerox PARCTab (Want et al., 1995) and PEPYS (Newman et al., 1991) projects, all are examples of office based, mobile, location-aware, applications.

Active Badge

The Active Badge project stemmed from Olivetti research laboratories in the early 1990's, it was an automated system for forwarding calls directly to employees, depending on their current location in the workplace. Each employee carried around a small personalised device, (the Active Badge) which using a transmitted IR-signal uniquely identified the carrier, the system could then track individual employee movements throughout the day and determine the closest available phone at any time.

ParcTab

The Xerox ParcTab centred on small wireless palm-sized hand held computers which where designed to host an office based application. The system provided a number of applications, not all of them context aware. Early attempts at context-aware applications like locating local resources or personnel, and obtaining room information, where merged with more conventional experiments with mobile office applications such as e-mail, calendar information or even weather forecasts.

Context-Aware frameworks

A significant challenge for context-aware systems lies in how we represent and process contextual data, this difficulty in dealing with low level data that drives context-based systems is a key factor in the relatively slow emergence of context-aware applications, a point emphasised by Pascoe, who stated that there is a “lack of a supporting infrastructure for capturing and processing context data, so hampering development of context-aware systems”. (Pascoe, 1997) So how do we represent contextual data?

With conventional development paradigms programmers use rigid structure and definite processes to control data, contextual data though is altogether more vague. This inherent uncertainty in contextual data, compared with a developer’s need to define static structure and rules, mean that we run the risk of compromising the contextual data itself by attempting to fit it into conventional developmental practises. Abowd sums this up by staying that “without good representations for context, application developers are left to develop ad-hoc and limited schemes for storing and manipulating key information” (Abowd, 1999). Therefore a wholly new novel approach is required for context-aware applications, an example of some of these frameworks include:

Stick-e Note

The Stick-e note architecture (Brown, 1996) (Pascoe, 1997) is a flexible, extensible, framework for handling context and related events, focusing on using context to trigger context related behaviours. Stick-e notes introduces the concept of attaching objects to contexts, much like Post-It^™ notes, the idea being that when a user enters such a context the object will be invoked. Each of these stick-e note objects is a link between a context (e.g. a place, person, time) and its content (e.g. information, action), along with triggering information. This triggering information defines the conditions under which the content should be invoked, for example we could attach a “go home message”, if the time is after 5pm.

Context Toolkit

The Context Toolkit (Dey, 2002) is a framework for a distributed system composed of context widgets. Context widgets, much like the idea of objects in object orientated approaches, are software components that provide functions for accessing contextual information, whilst hiding the complexity of context sensing. The toolkit provides services for encapsulation of sensors, access to context data through a network API, abstraction of context data, sharing of context data through a distributed infrastructure, storage of context data and privacy protection

CoBrA

CoBrA (Context Broker Architecture) (Chen, 2004) is an agent-based architecture which introduces the notion of intelligent spaces for supporting context-aware computing. Intelligent spaces are physical spaces (e.g., kitchens, cars, work office and meeting rooms) that contain intelligent systems that provide services to end-users. Pivotal to CoBrA is the idea of a context broker that centrally manages the agents. Agents can be either applications hosted on mobile devices or services provided by devices within rooms.

Location Awareness

The Xerox PARCTab and Active Badge projects are, like the majority of existing context-aware systems, examples of location-aware applications, that is they primarily use locational information in determining what services to provide the user. This is due to the fact that “the location of a device is an element of context that currently can be easily measured.” (Kassinen, 2003) Kassinen also identifies that there is a significant “user need for location-aware services.” (Kassinen, 2003) Context-aware systems dealing with locational information are widespread and the demand for them is growing due to the increasing spread of mobile devices (Baludauf et al, 2005) There are examples of location-aware systems in application areas such as GPS based navigation, tourist guidance (Berderson, 1995), supply-chain management (Kourouthanasis, 2002), and office tracking systems (Want et al., 1992) (Want et al., 1995).

As discussed earlier, one of the key challenges for a context-aware system is capturing contextual data (Pascoe, 1997) and for a location-aware system this involves the physical process of locating the device. There are a number of technologies available for collecting this positional data, the most significant for mobile devices include:

GPS

GPS (Global Positioning System) is a satellite navigation system, which can very precisely determine the position of a device, almost anywhere on the planet. The system is accurate to within 20 meters, although it is possible to greatly increase this precision through error-correction and differential GPS techniques. GPS devices do not work indoors or in some urban regions. Common examples of GPS driven devices technology include car or vehicle tracking systems, although GPS technology is more commonly integrated into modern mobile phones and PDAs.

Bluetooth™

Bluetooth™ is an international standard, which allows for wireless communication between enabled devices over a short range radio frequency. Bluetooth™ technology is both low cost and globally available, and is becoming increasingly common in small devices such as mobile phones and PDAs. It has a range of between 1 –100 meters, depending on the class of the device. Detecting Bluetooth™ devices requires specific hardware, and the location of a device is only available when the device is within range of the service area. There are also some concerns over privacy, blue jacking and bluesnarfing, respectively spamming and hacking discoverable devices, are examples of technology specific problems.

WLAN

WLAN (Wireless Local Area Network) was introduced as a wireless alternative to LAN, a method of networking devices. Although, originally prohibitively expensive, reducing costs, as well as the introduction of the IEEE 802.11a and g standards, has meant that many devices including laptops, printers, and PDAs now come with wireless capabilities as standard. A significant problem associated with WLAN is security, although this issue has been somewhat lessened since the introduction of encryption techniques WEP and WPA, and router based security measures such as MAC address screening. Like Bluetooth, WLAN requires specific hardware to locate devices, and the device is only available whilst in service range.

Mobile Context-Aware Systems

Mobile computing has been one of the largest computing related growth industries over the last decade, (Government Statistics, 2003) UK government statistics estimated that in 2003 73% of the adult population own or use a mobile phone. Mobile devices have saturated our everyday lives, placing small, powerful, programmable, computing devices into our pockets; ideal conditions for the first stages of ubiquitous computing. It is this integration that drives the need for context-aware systems specifically designed for mobile environments, which provide programs and services that react specifically to their current location, time and other environmental attributes. (Baldauf et al, 2005)

This highlights a need for context-aware applications for mobile devices, although existing frameworks are not specialised enough to support large scale mobile systems properly. Context-aware systems for mobile devices require more specialised architecture than conventional approaches. Mobile devices are continually moving and are not permanently connected to a network. This discontinuity of service has to be taken into account when designing architecture for mobile devices. (Tandler, 2001) Hofer et al say that Context-Awareness is especially interesting in mobile devices where the context of the application is highly dynamic (Hofer et al., 2005) Mobile devices are heterogeneous, so resolution, memory, computational power and network capacity all vary from device to device so specific considerations are required. (Hofer et al., 2003)

Hofer et al went further in identifying a number of requirements necessary for supporting context-awareness on mobile devices, these being:

• Context-sharing – the ability to share contextual data with other devices
• Meta information – must provide information about the conditions under which the contextual data has been deceived, i.e. the device from the sensor, its preciseness.
• Robustness – the architecture must be robust enough to handle disconnections from remote sensors.
• Extensibility – the architecture should support connections to multiple sensors
• Lightweightness – the architecture has to take into consideration restrictions of computational power.

(Hofer et al., 2003)

Mobile Context-Aware Frameworks

Hydrogen

The Hydrogen project (Hofer et al., 2002) is a specialised framework for mobile devices. Unlike most context-aware systems it relies less on the availability of a centralised component in distributed systems. Importantly, it distinguishes between remote and local context. Remote context being information another system knows about, and local context is information native to the system. Hydrogen provides mechanisms for sharing of this contextual data between devices in proximity, known as context sharing, using peer-to-peer technologies such as Bluetooth or WLAN.

Mobile Context-Aware Applications

B-MAD

B-MAD (Bluetooth Mobile Advertising) was introduced as a “system for delivering permission-based location-aware mobile advertisements to mobile phones, using Bluetooth™ positioning and WAP push delivery”. (Aalto et al, 2004) The system uses a Bluetooth sensor to pro-actively search for nearby enabled devices. Once located it obtains the devices unique address and passes this to an Ad Server. The Ad server, a PHP and MySQL platform, creates an advertisement in either XHTML or WML (Wireless Mark-up Language) depending on the capabilities of the device. Using a Push Sender the advertisement is then pushed back to the mobile device.

Mobile Based Advertising

Consider the following scenario, “Bob is shopping in a small mall with his Bluetooth enabled mobile phone, invisible to Bob his phone has been located by the application server and a shopping profile (containing relevant shopping related contextual data only) has been exchanged. Using this contextual data the system quickly matches Bob's details against the retailer’s product database and attempts to push an advertisement to his mobile device. Bob, a keen walker, receives the advertisement in the form of a WML message, offering him 10% off a pair of walking boots in his size.” This is a somewhat simplistic example, but it does highlight the distinct potential for context-aware mobile based advertising.

Due to the personal nature of mobile devices, pushing badly matched or unwanted information onto users is undesirable. (Aalto et al., 2002)(Kassinen, 2003) Aalto et al identified the importance of directly “targeting the advertisements, so the user is not drowned in adverts” (Aalto et al., 2002) and Kassinen indicates that most users don't mind being pushed information, as long as they really need it. Although, this is a somewhat vague notion, as what factors determine whether a user really needs a specific piece of information?

Aalto et al identified that the AD Server itself “has limited profiling capabilities”, and that there is a perceived need for profiling, so that “the user is not drowned in adverts”. This highlights the necessity for a more adaptive approach to more closely matching consumers with products.

Personalisation

Aalto et al identified that there is a need for personalisation of content, whilst talking about pushing advertisements to mobile devices. We extend this idea to embrace the concept of using contextual data obtained from a mobile device and using this to produce a completely personalised match. Effective personalisation would allow us to market products directly at an individual, as opposed to targeting a demographic.

We can look at personalisation as both the desire to more closely match customer details against product details and the ability to allow users to determine their preferences, i.e. we would like more of this type of product or no more of this type of product. An ideal situation would be to present a solution that provides a combination of these approaches, using contextual data to make an intuitive and learned best guess, whilst allowing users to indicate their feelings about the match and adapting accordingly. Lee et al introduced a system that learns user preferences through a user agent, and then uses a matching agent to closely match users to available services/events. (Lee et al., 2004)

In order to personalise services, we need to assume that device users rarely change, a point raised by Hofer et al. (Hofer et al., 2003) A scenario with regular changing users would drastically reduce the effectiveness of any personalisation.

Efficient Matching

Context-aware systems are by their very nature adaptive and because we cannot guarantee continuity of service (Tandler, 2001), as users are permanently moving in and out of service areas, this places considerable time constraints on the services provided by the system. Aalto et al identified that on average it took 37 seconds (for positioning and push latency) to push an advert to a mobile device using Bluetooth. (Aalto et al, 2002) As to whether this overhead was due to inefficiencies in the position mechanism or limitations in Bluetooth™ itself was not apparent, yet the implications clearly indicate that for mobile advertising systems to be viable, this overhead would need to be significantly reduced. However, with future developments it is likely that this time will significantly decrease, but it highlights that the decision process itself needs to be kept to an absolute minimum. The challenge here lies in producing high quality matches, which users will perceive as required or suitable, in a small amount of time.

Traditional matching algorithms are not wholly suitable to context-aware systems because contextual data is not always tractable, certain or easily matchable.

Unsolicited Advertising

Varshney et al suggested that mobile advertising is going to be a very important class of mobile commerce (Varshney et al 2002), but because of the personal nature of mobile devices there is broad scope for abuse. Using local connection technologies like Bluetooth or WLAN would provide a cost-free mechanism for pushing advertisements to end-users, this raises the issue of dealing with unsolicited advertisements or unwanted services, much like spam with e-mail. Mobile phone spam, in the form of unsolicited txt messages, has already reached epidemic proportions with 8 out of 10 mobile phones users reporting unsolicited messages (Jaques, 2005).

The key question here is how do we prevent unsolicited advertisements or unwanted services without interfering with legitimate ones or indirectly restricting the natural data flow of mobile context-aware systems.

The B-MAD system (Aalto et al., 2002) side steps the issue by stating that its framework focuses solely on permission-based advertising, therefore ruling out unsolicited advertising (i.e. spamming). Sadly, no mechanism for supporting such a permission based system was presented. Alternatively, it would be possible to consider other approaches derived from conventional spam prevention techniques, such as Bayesian filtering or service blocking techniques. A more likely possibility is to use contextual data to filter or restrict pushed advertisements based upon personal preferences and end-user habits. For example, we could define rules to only accept pushed adverts from shops that the end-user has purchased items from in the last year.

Privacy

Context can contain data that is sensitive, such as a user’s current location or current activity, so it is important to include some mechanisms for privacy. A view supported by Kourouthanasis et al says that “as context-aware system offer increased personalisation, privacy concerns for the capture and processing of context information is required.” (Kourouthanasis et al., 2002) Hong & Landay, go as far as saying that privacy is the most often-cited criticism of ubiquitous computing, and may be the greatest barrier to its long term success. (Hong & Landay, 2004) This point may seem alarmist, but end users ultimately reject applications they are uncomfortable using or find intrusive. (Hong & Landay, 2004) Because of the sharing nature of ubiquitous systems, we need to ensure that privacy and security concerns are balanced against the sharing nature or ubiquitous systems. Much previous work on privacy tended to focus on providing anonymity or on keeping personal information secret from hackers or unwanted sources.” (Hong, 2004) Therefore, it is essential that privacy concerns are included at every stage of development of context-aware applications, and “not done in an ad hoc manner or as an after thought.” (Hong & Landay, 2004) This being is a major criticism of the B-MAD (Aalto et al., 2002) system, which by admission has little or no consideration for privacy.

Existing systems for sharing contextual data such as Bluetooth™ and WLAN have previously had security concerns, but because of the surge in popularity of these technologies, emphasis has been placed on bolstering any security flaws, resulting in effective secure solutions.

A number of existing context-aware frameworks incorporate facilities for privacy, the Context Toolkit () introduces the concept of context ownership and CoBrA (Chen, 2004) has its own specific privacy policy language called Rei (Kagal et al., 2003). Alternatively, Confab (Hong & Landay, 2004) is an infrastructure aimed at simplifying the task of creating privacy-sensitive context-aware applications; it does this by addressing five important common features that need to be supported.

1. It identified three basic interaction patterns for privacy-sensitive applications, these being:
• Optimistic - applications share personal information and detect abuses
• Pessimistic - where more emphasis is placed on detecting abuses.
• Mixed-initiative - decisions to share information are made by the end-user.

2. Supports the tagging of personal information, that is marking personal information with personal preferences, for example, a person’s bank account may be available to certain secure institutions and not to anyone else.

3. Controls the access, flow and retention of personal information, allowing us to include restrictions based on context types, e.g. only allow co-workers to view my location between the hours of 9 to 5 pm.

4. Control over precision of disclosed data, e.g. we could disclose my address as either“16 Monks Road, Exeter” or “Devon”, depending of the source of the request.

5. Provides full logging facilities for both client and server, providing the ability to fully monitor the flow and use of data in the system.

Conclusion

Ubiquitous computing has enormous potential for changing the way we think about computing, although the field is still very much in its infancy. This is due to a couple of reasons, firstly, creating ubiquitous environments was expensive, most early work in ubiquitous computing originated from experimental in-house systems where large companies could foot the bill for creating a suitable environment. Secondly, traditional developmental frameworks are not designed for ubiquitous computing’s reliance on concurrency and parallelism; there was also a distinct lack of any supporting infrastructure. Although the sudden growth in popularity and availability of mobile devices has presented ideal conditions for ubiquitous computing and the emergence of projects like Indus presents a significant leap forward, the area is still very much embryonic. Ubiquitous computing though is here to stay

Context-aware applications offer a wealth of potential, and it will not be long before we start seeing a significant uptake in interest. When personal computers where introduced in the early 80's they radically altered our view of computing, giving rise to "desktop" computing, technological and user expectations changed, old dominant companies who didn't adapt quickly lost ground, new dominant companies appeared, computing fundamentally changed. As Weiser pointed out, “we still require much new work in operating systems, user interfaces, networks, wireless, displays and many other related areas.” (Weiser, 1994) This signifies some idea of the size of the challenge in hand. Ubiquitous computing though is likely to be an altogether quieter revolution.

The majority of early work on ubiquitous computing focused on context-aware applications, such as the Parctab and Active Badge projects, where the focus was not on producing commercially viable products. Whilst technically successful these projects where greeted with scepticism by employees, some where concerned over privacy, and others felt that the system was intrusive or that logged data would be misused. Highlighting that one of the biggest factors in a context-aware system is not necessarily the technical challenge but getting the end-user to accept the system.

Context-aware systems have as yet been limited to applications, which primarily use time or location as contextual information and most examples are experimental and non-commercial products. The success of GPS based systems though has proved that context-aware applications can be commercially viable products, although this success has only been possible because of the US’s decision to freely allow access to GPS technology and resources. Context-aware systems are unlikely to become widespread until adequate standardised methods of handling contextual data are commonplace. An illustration of this difficulty with handling contextual data being, how do we handle levels of abstraction in context-based systems. For example, take the simple question, “Where do I live?” This could provide numerous results, 16 Monks Road, Exeter, Devon, England, UK or Earth. Each of these answers is correct but usefulness depends on the context of the question. In terms of locating mobile devices the maturity of Bluetooth and WLAN technologies has presented ideal cheap, secure, globally available, mechanisms for local communication between devices.

Context-Aware applications for mobile devices present probably the biggest commercial opportunity of any ubiquitous systems, again technologies are in there infancy but it is unlikely that this key market will be ignored for long. Novel approach, like the B_MAD system (Aalto et al., 2002) have enormous potential but still have many hurdles (both technical and personal) to overcome in order to produce successful viable commercial solutions. As well as facing significant technical challenges, application developers face having to place a larger emphasis on more human factors. Marketing and advertisers are likely to have already seen the potential that mobile devices offer, at present advertisers focus their campaigns on particular demographics, but context-aware applications would allow them to directly target individuals.

For mobile-based context-aware systems to be accepted by the general public it is imperative that proper privacy and security measures are deemed to be in place, consumer confidence is key. End-users are already fearful of “big brother” scenarios, this was evident from projects like Active Badge, where the majority of problems where not technical but getting end-users to accept such systems. Products that seem intrusive or feel unsafe will be rejected; ultimately it is users who will shape the future.

References

Aalto, L., Gothlin, N., Korhonen, J., Ojala, T., Bluetooth and WAP push based location-aware mobile advertising system, MobiSys '04: In Proceedings of the 2nd international conference on Mobile systems, applications, and services, ACM Press, 2004, pp. 49-58.

Abowd., G. D., Dey, A., K., Wood, A., Applying Dynamic Integration as a Software Infrastructure for Context-Aware Computing, ICSE, 1998

Abowd., G. D., Dey, A., K., Toward a Better Understanding of Context and Context-Awareness, GVU Technical Report, 1999.

Abowd, G., Mynatt, E., Ubiquitous Computing: Past, Present and Future, ACM Transactions on Computer-Human Interaction, 2000, pp. 29-58

Abowd, G. D., Mynatt, E. D., Rodden, T., The Human Experience, Pervasive Computing, 2002, Vol. 1, pp. 48 -57.

Adult mobile phone ownership or use: by age, 2001 and 2003: Social Trends 34, 2003, National Statistics Online, viewed 14^th December 2005, <http://www.statistics.gov.uk/STATBASE/ssdataset.asp?vlnk=7202>

Baldauf, M., Dustar, S., Rosenberg, F., A Survey of Context-Aware Systems, International Journal of Ad Hoc and Ubiquitous Computing, 2005, forthcoming, <www.citeseer.ist.psu.edu/baldauf04survey.html>

Baudel, T., Beaudouin-Lafon, M., Charade: Remote Control of Objects Using Free-hand gestures, Communications of ACM, July 1993, Vol. 36, No 7, pp. 28-35

Bederson, B., Augmented Reality: A prototype Automated Tour Guide, In Proceedings of ACM CHI’95, 1995, Vol. 2, pp. 210-211.

Borah, K., INDUS: A New Platform for Ubiquitous Computing, viewed 29^th November 2005, <http://www.acm.org/ubiquity/views/v6i36_borah.html>

Brown, P., J., The Stick-e document: A framework for creating context-aware applications, In Proceedings of the Electronic Publishing, Palo Alto, 1996, pp. 259-272.

Chen, H., An Intelligent Broker Architecture for Pervasive Context-Aware Systems, 2004, PHD thesis, University of Maryland, Baltimore County

Dey, A., The Context Toolkit – a toolkit for context-aware applications, 2001, viewed 2^nd December 2005, <http://www.cc.gatech.edu/fce/contexttoolkit/>

Dey, A., Abowd, G., A Conceptual Framework and a Toolkit for Supporting the Rapid Prototyping of Context-Aware Applications, Human-Computer Interaction (HCI) Journal, 2001, Vol. 16, pp. 97-166.

Gelernter, D., Generative Communication in Linda, ACM Computing Surveys, 1985, Vol. 7, no. 1, pp. 80-112.

Global Positioning Devices, Wikipedia, viewed 15^th November 2005, <http://en.wikipedia.org/wiki/Gps>.

Hofer, T., Schwinger, W., Pichler, M., Leonhartsberger, G., Altmann, J., Context-awareness on mobile devices – the hydrogen approach, In Proceedings of the 36^th Annual Hawaii International Conference on System Sciences, 2002, pp. 292-302.

Hong, J., Landay, J., An Architecture for Privacy-Sensitive Ubiquitous Computing, MobiSys '04: Proceedings of the 2nd international conference on Mobile systems, applications, and Services, ACM Press, 2004, pp. 177-189.

Indus platform Developer, Aumeganetworks, 2005, viewed 10 December 2005, <http://developer.aumeganetworks.com/>.

Jaques, R., Spam plague hits 80 per cent of mobiles, www.computing.co.uk, 2005, viewed 29^th December 2005, <http://www.computing.co.uk/vnunet/news/2126701/spam-plague-hits-80-per-cent-mobiles>

Java API, java.sun.com, viewed 17^th November 2005, <http://java.sun.com/j2se/1.4.2/docs/api/index.html>

Kagal, L., Finin, T., Joshi, A., A policy language for a pervasive computing environment, In Proceedings of the 4^th IEEE International Workshop on Policies for Distributed Systems and Networks (POLICY), 2004, pp. 63-74.

Kassinen, E., User needs for Location-Aware Mobile Services, Personal and Ubiquitous Computing, 2003, Vol. 7, no.1, pp. 70-79.

Korkea-aho, M., Context-Aware Applications Survey, Internetworking Seminar (Tik-110.551), Helsinki University of Technology, 2000.

Kourouthanasis, P., Spinellis, D., Roussos, G., Giaglis, G., Intelligent cokes and diapers: MyGrocer Ubiquitous Computing Environment, First International Mobile Business Conference, July 2002, pp. 150-172.

Lee, G., Naganuma, T., Kurakake, S, Efficient Matching in a Context-Aware Event Notification System for Mobile Users, Massachusetts Institute of Technology, 2004, viewed 16 November 2005, <http://people.csail.mit.edu/gjl/research/publications/lee05efficient.pdf>

Newman, W., Eldridge, M., Lamming, M., PEPYS: Generating Autobiographies by Automatic Tracking, ECSCW '91, September 1991, pp. 175-188.

Picco, G. P., Balzarotte, D., Costa, P., LighTs: A Lightweight, Customizable Tuple Space Supporting Context-Aware Applications, SAC, 2005, pp. 413-419.

Pascoe, J., The Stick-e Note Architecture: extending the Interface Beyond the User, International Conference on Intelligent User Interface, 1997, ACM. pp. 261-264.

Scilit, B., Adams, N., Want, R., Context-Aware Computing Applications, IEEE Workshop on Mobile Computing Systems and Applications, 1994, pp. 88-90.

Tandler, P., Software Infrastructure for Ubiquitous Computing Environments: Supporting Synchronous Collaboration with Heterogeneous Devices, In Proceedings of UBICOMP2001, 2001, pp. 96-115.

Troung, K., Abowd, G., Botherton, J., Who, What, When, Where, How: Design Issues of Capture & Access Applications, In the Proceedings of UBICOMP 2001, 2001, pp. 209-224.

Varshney, U., Vetter, R., Mobile Commerce: Framework, Applications and Networking Support, Mobile Networks and Applications, 2002, Vol. 7, no. 3, pp. 185-198.

Want, R., Hopper, A., Falcao, V., Gibbons, J., The Active Badge Location System, ACM Transactions on Information Systems, 1992, Vol. 10, No 1, pp 91 – 102.

Want, R, Schilit, B., Adams, N., Gold, R., Petersen, K., Goldberg, D., Ellis, J., Weiser, M., The PARCTAB Ubiquitous Computing Experiment, IEEE Personal Communications, 1995, Vol. 2, no. 6, pp. 28-33.

Weiser, M., The Computer for the 21^st Century, Scientific American, 1991, pp. 94-10.

Weiser, M., Some Computer Science Problems in Ubiquitous Computing, Communications of the ACM, July 1993, pp. 137-143.

Weiser, M., The World Is Not a Desktop, ACM Interactions, Jan 1994, Vol. 1, no. 1, pp. 7 – 8.

Weiser, M., Ubiquitous Computing, 1996, Xerox, viewed 3^rd December 2005, <http://www.ubiq.com/hypertext/weiser/UbiHome.html>