Side-channel attacks exploit physical information leaked by cryptographic devices during computation, such as power consumption, electromagnetic emissions, or timing variations, to recover secret keys without breaking the mathematical algorithm. Power analysis attacks are the most common form for measuring the current drawn by a device during encryption and uses statistical methods or machine learning to correlate those measurements with the secret key. Deep learning-based side-channel analysis (DLSCA) has become the common approach in this field, where convolutional neural networks (CNNs) are used to learn and extract complex leakage patterns directly from raw power traces. This thesis presents a DLSCA attack targeting ASCON-128, the lightweight authenticated encryption algorithm selected by NIST as the standard for lightweight cryptography in 2023.