Identifying Spam Patterns in SMS using Genetic Programming Approach

SMS spam, also known as mobile spam, has become a prevalent and an ever growing issue due to the availability of bulk SMS services at nominal costs. These spam messages may not only be commercial but also pose a great deal of financial threats to the users. To fight against SMS spam, a variety of solutions have been proposed including content-based filtering, semantic indexing, machine learning classifiers, etc. However, in this regard evolutionary algorithms have not been utilized. Since the nature of SMS is contemporary, the representation of text messages keep evolving with the help of slangs, symbols, misspelled words, abbreviations and acronyms. Hence, such a solution is required which can accommodate these changes, also keeping the length of SMS in consideration. The model proposed in this paper generates regular expressions as individuals of population, using Genetic Programming Approach. These regular expressions so generated are used for the classification purpose. The application of Genetic Programming in the domain of SMS spam filtering has not been explored widely. It is able to eliminate False Positive errors, thus saving legitimate messages from being misclassified. The performance tends to improve with higher number of generations. Performance and confusion matrix for different number of generations are tabulated.