This study explores the enhancement of intrusion detection systems (IDSs) using deep learning and data augmentation techniques. The authors aim to address the limitations of traditional machine learning methods, particularly in handling imbalanced datasets, which often hinder the evaluation of model efficacy. Four prominent datasets—UNSW-NB15, 5G-NIDD, FLNET2023, and CIC-IDS-2017—are used to evaluate the performance of several deep learning architectures. The study focuses on a simple CNN-based architecture and more complex architectures like GRU and LSTM combined with CNN. Data augmentation techniques, such as synthetic minority oversampling technique (SMOTE), are applied to balance the datasets and improve model performance. The results show that the simple CNN architecture achieves high accuracy (up to 91% for the augmented CIC-IDS-2017 dataset) and that the quality and quantity of the dataset significantly influence the model's performance. The study highlights the potential of deep learning-based intrusion detection systems in enhancing cybersecurity frameworks by effectively detecting and mitigating sophisticated network attacks.This study explores the enhancement of intrusion detection systems (IDSs) using deep learning and data augmentation techniques. The authors aim to address the limitations of traditional machine learning methods, particularly in handling imbalanced datasets, which often hinder the evaluation of model efficacy. Four prominent datasets—UNSW-NB15, 5G-NIDD, FLNET2023, and CIC-IDS-2017—are used to evaluate the performance of several deep learning architectures. The study focuses on a simple CNN-based architecture and more complex architectures like GRU and LSTM combined with CNN. Data augmentation techniques, such as synthetic minority oversampling technique (SMOTE), are applied to balance the datasets and improve model performance. The results show that the simple CNN architecture achieves high accuracy (up to 91% for the augmented CIC-IDS-2017 dataset) and that the quality and quantity of the dataset significantly influence the model's performance. The study highlights the potential of deep learning-based intrusion detection systems in enhancing cybersecurity frameworks by effectively detecting and mitigating sophisticated network attacks.