Paying Attention to cyber-attacks: A multi-layer perceptron with self-attention mechanism

ABSTRACT

Cyber-attacks cause huge monetary losses to the institutions that are victims of them. Cyber-attack is becoming increasingly sophisticated. Therefore, the protection system against cyber-attacks has become a highly requested resource by any type of state or private institution. During the pandemic caused by COVID-19, the number of cyber-attacks against both public and private health institutions has increased. Cybersecurity systems have become a necessity. Various protection systems have been proposed using different machine learning algorithms, but deep learning consistently provides the best results. In this work we develop a deep learning model for the detection of different kinds of cyber-attacks, a study is carried out on the relevance of the selection of features in this type of algorithm and the importance of attention mechanisms is analyzed to improve the assessment of features within the same model. We have carried out the experiments using two datasets that are benchmarks in the field of cybersecurity and we have carried out a comparative study with both.