Classifying Web page features to detect DaaS

Abstract

A SaaS (Software as a Service) can transparently consume a DaaS (Data as a Service). However, heterogeneous DaaS and its evolution can disrupt the SaaS execution. In such cases, a middleware can provide such interoperability and monitor the DaaS trackback and its evolution by retrieving its metadata. For instance, the middleware MIDAS manually provides such interoperability. Considering the Web and the number of web pages and DaaS available, this task may be time-consuming and unfeasible. To automate this task, it is firstly important to distinguish a DaaS from a typical web page. Thus, this work aims to develop a model to identify DaaS from the Web. We collected a set of features from DaaS and non-DaaS pages to train our model, and we discuss some issues and strengths of our approach. We evaluate precision and recall, but we also measure the performance because this model will be embedded into a crawler in future versions of MIDAS. Our findings achieve high precision and low execution time, which can position our work in a proper direction to MIDAS evolution.