Abstract
Cybercrime is spreading rapidly as hackers find new ways to take advantage of flaws in the global information infrastructure. Ethical hackers place a greater emphasis on identifying security flaws and proposing solutions. In the area of cyber security, there is a significant need for the creation of efficient methods. The complex and unpredictable nature of cyber-attacks against communications networks indicates that most techniques employed in today's Intrusion Detection System (IDS) are inadequate. Machine learning (ML) in cyber security has lately gained prominence due to its efficiency in cyber security challenges. Intrusion, malware, spam, and phishing categorization and identification are just some of the primary challenges that ML approaches were applied to in the field of cyber security. While ML can't be used to automate a whole cyber security system, it does aid in the identification of cyber security threats more effectively than other software-oriented techniques, which in turn relieves pressure on security analysts. Data from NSL-KDD is collected and analysed in this paper. Correlation-based Feature Selection (CFS) and Information Gain (IG) are employed to pick out the most relevant aspects of the processed data. Random Forest (RF), K-Nearest Neighbour (KNN), and Naive Bayes (NB) are the three ML models created. In the end, the measurements are used to determine which method of ML model and feature selection is the most effective. Good accuracy rates decreased false rates, and manageable computational and communication expenses can all result from combining ML with an effective feature selection strategy.
Keyword
Cyber-attack, One Hot Encoding, Machine Learning, Feature Selection, Internet, Accuracy.
PDF Download (click here)
|