Um estudo de algoritmos de aprendizagem de máquinas para o Smart Defender
With the expansion of the Internet, coupled with the growing number of Internet of Things device devices (IoT), denial of service attacks (DoS), as well as their distributed variant (DDoS), It’s becoming a significant problem for the availability of services operating on the Internet. Thinking ab...
Na minha lista:
| Autor principal: | |
|---|---|
| Outros Autores: | |
| Formato: | bachelorThesis |
| Idioma: | pt_BR |
| Publicado em: |
Universidade Federal do Rio Grande do Norte
|
| Assuntos: | |
| Endereço do item: | https://repositorio.ufrn.br/handle/123456789/48196 |
| Tags: |
Adicionar Tag
Sem tags, seja o primeiro a adicionar uma tag!
|
| Resumo: | With the expansion of the Internet, coupled with the growing number of Internet of Things
device devices (IoT), denial of service attacks (DoS), as well as their distributed variant
(DDoS), It’s becoming a significant problem for the availability of services operating on the
Internet. Thinking about this, the researcher Francisco Sales de Lima Filho, proposed the
Smart Defender system, a distributed, non-invasive system (compatible with the current
network scenario) and with a collaborative approach, to be executed at all levels of providers,
aiming to overcome DoS / DDoS attacks as close as possible to their origin. The system
consists of the subsystems for detection (Smart Detection), protection (Smart Protectiont)
and monitoring (Smart Monitoring). This work aims to analyze the performance of machine
learning algorithms that can compose the core of the detection module. A study is done on
the Random Forest, Decision Tree, Logistic Regression and AdaBoost algorithms as well
as testing using the Python Scikit-Learn library to identify the best performing algorithm.
The database for use in performance tests was the database created by Sales for Smart
Defender research. With the tests carried out, the performance of the classifiers was verified
using the metrics of accuracy and Kappa index. At the end of the study, it was found that
AddaBoost has a slightly higher performance than the other algorithms. |
|---|