Poster Papers

Secure context-aware reconfiguration for mobile devices

AutorLeonid Batyuk, Sahin Albayrak
QuelleProceedings of the 5th Future Security Research Conference, Berlin 
LinksBibTeX 

Over the past ten years, the trend in user applications has definitely moved towards autonomous mobile devices, and wireless networks. More and more applications rely on various physical and logical sensors. The computing environments of today have become highly heterogeneous and spontaneous, but the models and methods for software development remained mostly the same as for desktop computers and wired networks. These models do not allow software to adapt to the current situation of the user. A novel approach to solving this problem is context-awareness, where software is enabled with knowledge about its environment and adapts itself to the current needs of the user. During the last decade, numerous context-awareness frameworks have evolved on both infrastructure-level and device-level. Most context-aware systems rely on predefined settings which invoke a certain action upon a change in the environment. However, in several works an attempt has been made to overcome the fixed model and implement a generic approach. The concepts of context proximity and context familiarity have been introduced, which, to some extent, enable reasoning upon context without having a precise set of rules and predefined conditions. These approaches provide more flexibility, but less control due to inability of such methods to annotate context information with semantic data, enabling full-fledged reasoning. As such, automated semantic recognition and annotation of context is a fundamental problem we address. Utilizing contextual information for autonomous secure self-configuration is an example of a use case where manual rule-base management is unfeasible, but also current heuristics fail due to their shortcomings. We contribute to existing heuristic approaches with methods of artificial intelligence, utilizing techniques known from the field of anomaly detection to identify, predict and annotate context information, allowing to make more precise and reliable security decisions upon sensor data.