This study proposes a deep learning and data augmentation approach to enhance intrusion detection systems (IDSs). The research addresses the limitations of traditional machine learning methods in intrusion detection, particularly the challenges posed by imbalanced datasets. The study evaluates several deep learning architectures on four prominent datasets: UNSW-NB15, 5G-NIDD, FLNET2023, and CIC-IDS-2017. Data augmentation techniques, such as SMOTE, are applied to balance the datasets and improve model performance. The results show that a simple CNN-based architecture achieves high accuracy (up to 91% on the augmented CIC-IDS-2017 dataset), while more complex architectures show only marginal improvements. The study highlights the effectiveness of deep learning-based intrusion detection in enhancing cybersecurity frameworks by detecting and mitigating sophisticated network attacks. The findings indicate that the quality and quantity of the dataset significantly influence model performance. The study also demonstrates that data augmentation is crucial for fair evaluation of deep learning models in intrusion detection. The results suggest that simple CNN-based models with data augmentation can achieve high accuracy, making them suitable for intrusion detection tasks. The study concludes that deep learning models are effective in capturing complex patterns in network traffic and can be used to improve intrusion detection systems. The research contributes to the field of cybersecurity by providing a practical approach to enhance intrusion detection systems using deep learning and data augmentation.This study proposes a deep learning and data augmentation approach to enhance intrusion detection systems (IDSs). The research addresses the limitations of traditional machine learning methods in intrusion detection, particularly the challenges posed by imbalanced datasets. The study evaluates several deep learning architectures on four prominent datasets: UNSW-NB15, 5G-NIDD, FLNET2023, and CIC-IDS-2017. Data augmentation techniques, such as SMOTE, are applied to balance the datasets and improve model performance. The results show that a simple CNN-based architecture achieves high accuracy (up to 91% on the augmented CIC-IDS-2017 dataset), while more complex architectures show only marginal improvements. The study highlights the effectiveness of deep learning-based intrusion detection in enhancing cybersecurity frameworks by detecting and mitigating sophisticated network attacks. The findings indicate that the quality and quantity of the dataset significantly influence model performance. The study also demonstrates that data augmentation is crucial for fair evaluation of deep learning models in intrusion detection. The results suggest that simple CNN-based models with data augmentation can achieve high accuracy, making them suitable for intrusion detection tasks. The study concludes that deep learning models are effective in capturing complex patterns in network traffic and can be used to improve intrusion detection systems. The research contributes to the field of cybersecurity by providing a practical approach to enhance intrusion detection systems using deep learning and data augmentation.