ANALYSIS OF METRICS TO EVALUATE SYNTHETIC NETWORK MONITORING DATA

• Grant number: 24/22752-9
• Support Opportunities: Scholarships abroad - Research Internship - Scientific Initiation
• Start date: April 01, 2025
• End date: July 31, 2025
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computer Systems
• Principal Investigator: Magnos Martinello
• Grantee: Guilherme Silveira Gomes Brotto
• Supervisor: Idilio Drago
• Host Institution: Centro Tecnológico. Universidade Federal do Espírito Santo (UFES). Ministério da Educação (Brasil). Vitória , SP, Brazil
• Institution abroad: Università degli Studi di Torino (UNITO), Italy
• Associated to the scholarship: 23/11477-4 - ANALYSIS OF METRICS FOR EVALUATION OF SYNTHETIC NETWORK MONITORING DATA, BP.IC

Abstract

Several applications and research involving the area of ¿¿computer networks require datasets to be carried out. However, due to factors such as privacy laws or internal company policies, access to these datasets may be limited. In this sense, studies that aim to make this type of data available are essential. An alternative to this problem is to use models that can generate synthetic data from real data, preserving their characteristics and maintaining privacy. In this context, in a scientific initiation project conducted within the research group, we investigated how the generation and evaluation of some types of mobility data can be done efficiently using optimized models for time series, particularly in the generation of attack data in IoT environments. Although efficient, the models investigated so far have limitations regarding the properties of the data that they are capable of learning. Therefore, this research project aims to investigate and analyze metrics that quantify the quality of data generated from generative models, especially those generated from generative adversarial networks, stable diffusion and transformers. To this end, analyses and characterization of synthetic data will be performed, generated from recent models in the literature, using real datasets as input.