SMS spam filtering: Methods and data

Abstract

Mobile or SMS spam is a real and growing problem primarily due to the availability of very cheap bulk pre-pay SMS packages and the fact that SMS engenders higher response rates as it is a trusted and personal service. SMS spam filtering is a relatively new task which inherits many issues and solutions from email spam filtering. However it poses its own specific challenges. This paper motivates work on filtering SMS spam and reviews recent developments in SMS spam filtering. The paper also discusses the issues with data collection and availability for furthering research in this area, analyses a large corpus of SMS spam, and provides some initial benchmark results.

SMS Spam Filtering Using Probabilistic Topic Modelling and Stacked Denoising Autoencoder

“…This is by no means a unique problem for spam filtering, however, the limited available text per SMS makes the feature space sparse. This means that the samples, from the input space, are fewer and further apart, thus significantly reducing the data that the classifier has to work with [5]. Hidalgo et al [6] suggested the use of different features including: normalised words, character bi-and tri-grams and word bi-grams.…”

“…However, they argue that the greatest disadvantage of this approach was that the features are fixed and can not be updated as the data changes and the nature of spam threat changes. The extracted features tend to be high dimensional requiring some sort of feature selection, or dimensionality reduction techniques [5,15,6].…”

“…Prior research has shown that the most effective approach for spam filtering is to perform the threat analysis on the message content level [5]. The SMS problem is in principle very similar to email spam filtering [8,2].…”

“…Despite this most filters use standard feature extraction methods such as direct N-gram characterbased and word-based tokenisation [6]. Supervised and unsupervised machine learning techniques are commonly trained using a collection of labelled messages of spam and non-spam (usually referred to as ham) [5]. The trained model is then used to predict labels of previously unseen messages.…”

Biterm for spam filtering in short message service text

“…Spam messages may fill user's mailbox engulfing important personal mails, wasting network bandwidth, consuming user's time and energy to sort through it [4]. Similarly, SMS spam messages are also nuisance to mobile subscribers as many mobile subscribers have suffered financial losses resulting from responding to SMS spam senders on premium rate numbers and signing up to expensive subscription services [5], [6].…”