In this article, we develop a comprehensive mobile-based approach, which is able to perform the essential processes needed to automatically analyze and detect epileptic seizures using the information contained in electroencephalography (EEG) signals. We first develop and implement an appropriate combination of different algorithms that resample, smooth, remove artifacts, and constantly and adaptively segment signals to prepare them for further processing. We then improve and fully implement a large variety of features introduced in the literature of epileptic seizures detection. We also select the relevant features to reduce a feature vector space and improve the classification process by developing two automated filter and wrapper selection algorithms. We thoroughly compare between these selection algorithms in terms of redundant features, execution time and classification accuracy through three experiments. We subsequently exploit the selected features as input to a machine learning classifier to detect epileptic seizure states in a reasonable time. We experimentally and theoretically evaluate the scalability of the whole algorithm respectively on patients' data available in standard clinical database and on 500 EEG recordings including 500 seizures. Having efficient and scabble algorithm, we develop two extra algorithms to dynamically acquire and transmit EEG signals from wireless sensors attached to patients and to visualize on mobile devices the obtained processing and analysis results. We finally integrate all our algorithms together along with an android mobile application to implement an effective mobile-based EEG monitoring system where its accuracy is tested on live EEG data. (C) 2015 Elsevier Ltd. All rights reserved.