Study and Experimental Analysis of Metaheuristic Based Optimizers with respect to P2P Botnet Detection

Cybercrimes are growing at an exponential rate everyday and has become a threat to both people and corporations. Businesses and government organizations that entirely rely on digital information are more prone to the cyber attacks and should construct defense mechanism against them promptly, before they can cause any consequential damage. Malware software such as viruses, worms, spyware, Trojan horses, botnets ,etc. are often used in cybercrimes. In recent times, botnets have gained a name for being used in majority of internet attacks. Botnets are a network of compromised computers that attacker secretly controls and utilizes for harmful purposes. Botnets are the most innovative of all digital assaults because they are difficult to identify. The botnets can be classified into two groups based on their nature of propagation; centralized botnets and decentralized or peer - to- peer botnets. The peer-to-peer botnets has emerged as a new advance form of botnet which is extremely difficult to detect. The peer-to-peer botnets are very strong against defense countermeasures than the conventional centralized botnets. In this paper, three supervised machine learning classification techniques namely decision tree, random forest and ensemble learning technique are implemented and analyzed for p2p botnet attack and their performance are evaluated based on accuracy, precision, recall and f1-score. Along with this, study and experimental analysis of meta-heuristic based optimizers are being implemented with respect to P2P botnet detection.