Abstract
The Internet of Things work by constantly sensing the physical properties in the vicinity of the user such as ambient light, sounds, motion and temperature. These sensors produce huge volumes of data that has to be efficiently sifted for relevant events required triggering certain actions. In addition, filtering has to be performed to ensure that privacy-sensitive confidential data is not leaked. Efficient and expressive pattern matching is thus a key enabling technology for the full realization of ambient and humanized computing. The bulk of research in this area has focused on the use of specialized hardware and reducing of the memory footprint. Unfortunately, there has been limited work if any on optimizing the core elements of pattern matching- the regular expression language and the compilation process that is responsible for converting patterns into internal data structures. The importance of writing good REs so that on compilation they do not lead to unrealizable data structures is relatively less understood. In the proposed research, we empirically compare different RE processing engines and practically demonstrate that the compilation phase is highly memory intensive and time-consuming as compared to the matching phase -and hence is worth exploring for new techniques and optimizations. As a second important contribution, we propose a novel technique for defining regular expressions by utilizing JavaScript Object Notation. Our evaluation with carefully created patterns shows that the performance of the proposed technique is at par with competing approaches. It is also less ambiguous, extensible, more expressive and much appropriate for defining large and complex patterns.