| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| DM_PedroSampaio_MEI_2020 | 4.03 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
As aplicações Web têm um papel fulcral em muitas organizações, suportando e potenciando os seus negócios. A sua popularidade e criticidade é também apelativa aos atores maliciosos que exploram possíveis vulnerabilidades para executar os seus ataques, levando a prejuízos significativos de reputação e financeiros por partes das organizações e pessoas afetadas.
Neste projeto explorámos a utilização de algoritmos de Machine Learning para detetar anomalias e possíveis ciberataques através da análise dos logs que a aplicação produz, com base em dados reais e dados gerados a partir de testes de intrusão foi criado um dataset para análise através de algoritmos Machine Learning de modo a determinar ciberataques em curso. O dataset é composto por 36 parâmetros resultantes da agregação de dados por janela temporal e do enriquecimento de dados. Os datasets foram tratados e analisados através de diferentes algoritmos de Machine Learning considerando a atualização de híper parâmetros verificando se a existência de algoritmos capazes de detetar anomalias e possíveis ciberataques com
uma taxa de acerto superior a 90%. A eficácia verificada contribui para a criação de uma solução independente focada no desempenho e na rapidez de resposta que permite detetar e/ou bloquear ataques numa fase inicial mitigando as suas consequências negativas junto das organizações.
Web applications play a central role in many organizations, supporting and enhancing their business. Its popularity and criticality is also appealing to malicious actors who exploit possible vulnerabilities to carry out their attacks, leading to significant reputational and financial losses by affected organizations and people. In this project we explored the use of Machine Learning algorithms to detect anomalies and possible cyberattacks through the analysis of the logs that the application produces, based on real data and data generated from intrusion tests, a dataset was created for analysis using Machine Learning algorithms in order to determine ongoing cyberattacks. The dataset is composed of 36 parameters resulting from the aggregation of data by time window and the enrichment of data. The datasets were treated and analysed using different Machine Learning algorithms considering the update of hyperparameters verifying the existence of algorithms capable of detecting anomalies and possible cyberattacks with a hit rate greater than 90%. The verified effectiveness contributes to the creation of an independent solution focused on the performance and speed of response that allows detecting and/or blocking attacks at an early stage, mitigating their negative consequences for organizations.
Web applications play a central role in many organizations, supporting and enhancing their business. Its popularity and criticality is also appealing to malicious actors who exploit possible vulnerabilities to carry out their attacks, leading to significant reputational and financial losses by affected organizations and people. In this project we explored the use of Machine Learning algorithms to detect anomalies and possible cyberattacks through the analysis of the logs that the application produces, based on real data and data generated from intrusion tests, a dataset was created for analysis using Machine Learning algorithms in order to determine ongoing cyberattacks. The dataset is composed of 36 parameters resulting from the aggregation of data by time window and the enrichment of data. The datasets were treated and analysed using different Machine Learning algorithms considering the update of hyperparameters verifying the existence of algorithms capable of detecting anomalies and possible cyberattacks with a hit rate greater than 90%. The verified effectiveness contributes to the creation of an independent solution focused on the performance and speed of response that allows detecting and/or blocking attacks at an early stage, mitigating their negative consequences for organizations.
Description
Dissertação de Mestrado em Engenharia Informática
Keywords
Machine Learning ciberataques algoritmos atividade maliciosa
Pedagogical Context
Citation
Publisher
Instituto Politécnico do Porto. Escola Superior de Tecnologia e Gestão