Hybrid big bang-big crunch with ant colony optimization for email spam detection

Electronic mails (emails) have been widely adapted by organizations and individuals as efficient communication means. Despite the pervasiveness of alternate means like social networks, mobile SMS, electronic messages, etc. email users are continuously growing. The higher user growth attracts more spammers who send unsolicited emails to anonymous users. These spam emails may contain malware, misleading information, phishing links, etc. that can imperil the privacy of benign users. The paper proposes a self-adaptive hybrid algorithm of big bang-big crunch (BB-BC) with ant colony optimization (ACO) for email spam detection. The BB-BC algorithm is based on the physics-inspired evolution theory of the universe, and the collective interaction behavior of ants is the inspiration for the ACO algorithm. Here, the ant miner plus (AMP) variant of the ACO algorithm is adapted, a data mining variant efficient for the classification. The proposed hybrid algorithm ((HBC)-C-3-AMP) adapts the attributes of (BC)-C-3 (BB-BC) for local exploitation and AMP for global exploration. It evaluates the center of mass along with the consideration of pheromone value evaluated by the best ants to detect email spam efficiently. The experiments for the proposed (HBC)-C-3-AMP algorithm are conducted with the Ling Spam and CSDMC2010 datasets. Different experiments are conducted to determine the significance of the pre-processing modules, iterations, and population size on the proposed algorithm. The results are also evaluated for the AM (ant miner), AM2 (ant miner2), AM3 (ant miner3), and AMP algorithms. The performance comparison demonstrates that the proposed (HBC)-C-3-AMP algorithm is superior to the other techniques.

Automated summary

This paper proposes a self-adaptive hybrid algorithm of big bang-big crunch (BB-BC) with ant colony optimization (ACO) for email spam detection, and it demonstrates its superiority over other techniques through experiments.