Untitled CONTEXT AWARE SOFTWARE AND SYSTEMS

In the FIDIS deliverable D7.3 “Report on Actual and Possible Profiling Techniques in the Field of Ambient Intelligence” it is stated that profiling is one of the central techniques necessary to build and provide sophisticated AmI space and the schematic view below is given (see ):

Figure : Schematic view of profiling activity in the AmI environment

Although an important part - profiling is still only part of the background processing that makes AmI acting ‘intelligent’. A broader conceptual view on this is expressed by the techniques called context-awareness and context-aware computing.

First of all it needs to be clarified what the term context in the area of context-aware computing means. Actually a lot of different definitions exist which all cover different aspects of the concepts related to context in the field of computer science and computation. A general and intuitive definition is given in [Lieberman et al. (2000)] describing it as “… everything that affects the computation except the explicit input and output …”.

As refinement of this very general definition, approaches could be found which try to specify categories of information that set up the context. Examples of these categories could be the user’s location, social status or behaviour properties of the environment including capabilities of computing devices etc. (see also ). Some authors even try to build complex hierarchies of these classes of information [Schmidt et al. (1999)].

Nevertheless these ‘enumerative’ approaches have the disadvantage that they are very domain and application specific. To overcome this limitation a more conceptual definition of the term context was developed. The following definition taken from [Franz et al. (2007)] tries to unify these definitions and to establish a common notion for the term ‘context’:

Context is the state of an entity’s environment related to its activities. The part of the potentially available information which is considered to be relevant for influencing or triggering these activities will be described by a model.

The model defines 

1. which features are potentially considered to describe the state, 

2. whether these features are directly or indirectly derived from measurements provided by physical or virtual sensors, 

3. dependencies between these features, 

4. dependencies of features from former states, and 

5. how to draw conclusions about the state based on these features. 

An instance of this model describes the current state of the entity’s environment.

Given this definition, the authors of [Franz et al. (2007)] concluded that context-awareness could be defined as follows:

Context-Awareness means that activities of an entity are influenced or even triggered by knowledge of the current context. A characteristic feature of context-aware systems is the aim to support interactions between an entity and its environment.

As it could be seen from these definitions profiling and the environmental awareness combined with the related reactions as shown in are just some aspects (or a refinement) of context and context-awareness.

The following example may illustrate this: The foundation is the scenario of the AmI bar as described in the FIDIS deliverable D2.2. On the one side the AmI bar has to perform all kinds of profiling techniques to find out that a certain guest of the bar usually orders a certain type of beer. But knowing this profile alone is not sufficient for an ‘intelligent behaviour’ of the AmI bar. It needs more context information - especially which types of beer are in stock - to react appropriated using context-aware computing. Thus the profile is part of the context.

The FIDIS deliverable D7.3 discusses extensively the implications and risks coming from profiling and describes known solutions at technical, sociological and legal level. As context-aware computing can be understood as enhanced version of the profiling related processing techniques at least the same problems regarding identity, privacy and protection of personal data can be foreseen. 

Besides what is already described here, we want to concentrate on context-aware computing as emerging and enabling technology for privacy enhanced AmI space. As discussed in Section , in a privacy enhanced AmI space the fixed sensors of the environment should emit information about themselves. These huge amounts of unstructured data need to be filtered and pre-processed in order to inform (or even react on behalf of) the user in a non-annoying manner. Context-aware computing seems to be one key technology to achieve this goal. The personal device of the user (responsible for informing him) needs to be aware of the current situation, user’s preferences etc. (hence the relevant context) to make ‘intelligent’ decisions very similar to the way the surrounding AmI space makes its decisions.

Because AmI environments are somewhat speculative things of the future, the usage of context-aware computing to support PETs is illustrated below on the most elaborated example of collaborative eLearning.

Regarding privacy, collaborative eLearning is very contradictory: on the one side users want to work (learn) together e.g. by aiding each other, which usually requires a lot of information about the different participants. On the other side a user might not want to disclose too much information about himself, in order to avoid being associated with under achievement etc. This becomes especially more complicated if the user is a member of different learning groups. 

BluES’n is an example for such a collaborative eLearning environment, which is based on workspaces, in which users can learn and work together [Borcea-Pfitzmann et al. (2005)]. From a privacy point of view a partitioning regarding disclosure of information related to the different workspaces is necessary. This can be achieved by intra-application partitioning (IAP) as described in [Borcea et al. (2005)]. However IAP so far uses classical identity management technologies, which require the user to choose a particular partial identity for every relevant situation. Drawbacks of usability are the shady side of this because:

1. IAP must be applied additionally to the primary tasks the user wants to perform and 

2. The accruement of data and their influence on privacy are often not obvious to the user. 

Therefore context-aware computing is used to support the user. Especially two aspects are of importance: 

1. Raising of privacy awareness by informing the user continuously about his privacy state and 

2. Supporting of the usage of IAP by making automatic decision (or at least suggest them) based on the current context (which covers the user’s privacy preferences). 

Figure : Context-aware computing to support privacy - illustrated by the collaborative eLearning environment BluES’n [Franz et al. (2007)]

Within BluES’n a context-aware component called Decision Suggesting Module (DSM) is responsible for these tasks (see ). Particularly, the DSM has to evaluate the current context and to generate a suggestion (or to make an automatic decision) regarding which partial identity an initiated action should use [Franz et al. (2006)]. All information related to this action - data requested by the server due to access control policies, data explicitly sent by the user, but also information about the action itself - can be assigned to this partial identity. Additionally, relevant contextual information will be presented in a privacy aware user interface [Franz et al. (2006a)].

However, one also has to consider that users want to be recognised by others in a collaborative scenario to enable reasonable working. Thus, it is not possible to perform all actions anonymously or under different partial identities, respectively, to ensure their unlinkability. The DSM has the challenging task to support these two oppositional goals. As already said it is of course necessary that the user defines his preferences regarding these goals since this cannot be done automatically. 

New Shapes for Computational Devices

A number of trends in technological development will lead to new shapes for computational devices. In this context the most important ones seem to be: 

1. Miniaturisation of integrated circuits (ICs) 

2. Integration of more and more functions (e.g. increased computational power, communication using different types of networks, network protocols etc.) 

3. Development of new and smart materials 

Early examples are, for example, new shapes of RFID tags integrated into clothing. Far more advanced are wearable computers that integrate computational systems, interfaces and displays in clothes [Lisetti and Nasoz (2004)].  

Other examples for new shapes are mobile devices with more and more integrated functionality. Already today PDAs can be used as mobile phones, computers with office applications, web-surfing devices, and media players. Integration of additional functions will increase in future.  

Since 1997 Pister et al. carried out research at Berkley University with the target to develop self-contained, millimetre-scale sensing and wireless communication devices (so called motes) for massively distributed sensor networks. A number of these wireless connected motes are called Smart Dust. The working group of Pister suggested many different areas of use for Smart Dust, many of them with a military or secret service background. Recently, Steele (2005) summarised state-of-the-art in Smart Dust related research, development and application.

New shapes for computational devices will play a major role in the context of sensors, actuators and computational devices for AmI. They also may be used for advanced identity management devices, based on today’s PDAs. 

Research seems to be driven by public as well as private organisations, depending on the planned use of the resulting products. While mobile phones and PDAs currently are developed by private enterprises, in the area of Smart Dust research mainly seems to be driven by public research institutions.