Abstract
Due to proliferation of online posts and rise in the active social media users, fraudulent activities related with spam messages have taken a spike drift. Spam is an activity by which hackers use electronic messaging system to unsolicited messages in mass content to unknown users. It can be also taken as one of the major attraction of attackers in the form of short message service (SMS) messages. Spam messages can be categorized in different categories such as business opportunity spam, trending topic spam, banking services spam etc. These problems can be tackled by confirming to the actions taken by users towards these messages. There is an urgent explicit need of practical medium in order to assist the users against these spam messages. This paper proposes a novel SMS spam filtering model based on Danger theory of Artificial Immune System (AIS). Several feature extraction and selection techniques have been applied for optimizing the algorithm and claiming an admissible accuracy. This paper uses a biologically inspired algorithm named Krill herd Optimization (KHO) for the task of feature selection and various optimization functions like Quing function, Sumsquare function, Levy function etc. are applied for enhancing its performance. The Dendritic Cell Algorithm (DCA) is also incorporated with KHA as an added advantage towards achieving efficiency. Comparative results between Dendritic Cell Algorithm (DCA) with KHA and other spam filtering models have been shown in comparison with several state-of-the-art machine learning classifiers. The algorithms have been experimented by using varied optimization functions illustrated using visualization tools and results have been validated in the paper. The obtained results demonstrate an admissible accuracy of 96% that is calculated using different information retrieval metrics using recall, F-measure and precision.