A Lightweight Deep Neural Model for SMS Spam Detection
Wei Feng, Trang Nguyen
Abstract
The short messaging service (SMS) is one of the most popular and also most affordable telecommunication services. The popularity and affordability of SMS, however, have made it an ideal target for spamming. Spam is a major nuisance to mobile subscribers, but can also lead to security breaches or criminal activities. In this paper, we propose a novel lightweight deep neural model called Lightweight Gated Recurrent Unit (LGRU) for SMS spam detection. In addition, we incorporate enhancing semantics retrieved from external knowledge (WordNet) to augment the understanding of SMS text inputs for better classification. We compare the performance of our proposed model with that of more than 30 SMS spam classifiers that use various conventional machine learning and deep learning techniques. Experimental results show that our model outperforms these existing classifiers in terms of precision, recall and accuracy. In addition, our model requires fewer training parameters and incurs significantly less training time than state-of-the-art deep learning based classifiers.