Optimizing quantum machine learning for proactive cybersecurity

The evolution of cyberattacks requires a continuous race to implement increasingly sophisticated techniques that allow us to stay ahead of cybercriminals. Thus, a relevant inverse problem in cybersecurity involves determining underlying patterns or models of possible cyber threats based on observed data. In particular, the processing of massive data combined with the application of Machine Learning methods and other techniques derived from Artificial Intelligence have so far achieved very significant advances in preventing and mitigating the impact of many cyberattacks. Given that the keyword in cybersecurity is anticipation, this work explores the possibilities of quantum computing and, in particular, of Quantum Machine Learning to have, when the quantum computing era arrives, the most optimal parameterisations to put these models into practice. Although the application of quantum technologies in a real context may still seem distant, having studies to assess the future viability of Quantum Machine Learning to identify different types of cyberattacks may be a differential element when it comes to ensuring the cybersecurity of essential services. For this reason, this work aims to use several datasets of known problems in the field of cybersecurity to evaluate the most optimal parameterisations in some known Quantum Machine Learning models, comparing the results with those obtained using classical models. After analysing the results of this study, it can be concluded that Quantum Machine Learning techniques are promising in the context of cybersecurity, giving rise to future work on a wider range of cybersecurity datasets and Quantum Machine Learning algorithms.