Performance evaluation of machine learning algorithms for network anomaly detection: an approach through the AHP-TOPSIS-2N method
Gabrielle Barbosa do Nascimento, Marcos dos Santos · Procedia Computer Science · 2022
The biggest challenge for cyber security lies in the detection of anomalies in networks. Machine learning techniques participate in the automation of the detection process. In this scenario, this study aimed to evaluate the performance of machine learning algorithms used in the classification of network access. This evaluation has adopted the AHP-TOPSIS-2N hybrid multi-criteria decision support method. This approach allows a robust structuring of the problem by enabling decision-makers to assign weights to the criteria and by determining the score of the alternatives regarding a positive ideal solution. This approach generates a consistent analysis as it produces two orderings of algorithms based on their key performance metrics together. This provides greater safety in choosing the most suitable algorithm for the employment in the detection of network anomalies.