Abstract: In the digital age, the increasing sophistication and variety of cyber threats highlight the importance of strengthening cybersecurity to protect current networks. This study proposes an approach for the early detection of attacks, using the Matrix Profile (MP) technique to analyze network data streams as time series in an online manner. This method focuses on identifying anomalies in the network as early indicators of network attacks, addressing the limitations of existing Machine Learning systems that predominantly rely on offline training and struggle to recognize patterns of new or untrained attacks. Our proposal was evaluated in various attack scenarios, demonstrating superior performance metrics compared to traditional methods such as CUSUM, EWMA, and ARIMA.