Machine learning-based network intrusion detection systems (MNIDS) offer numerous advantages, including cost savings, monitoring, fast and accurate detection of DoS/DDoS attacks. One of the most critical aspects impacting the efficacy of this machine learning model is the used dataset in the machine learning models. However, even several N-IDS datasets have been developed, the greatest problem is data imbalance and a lack of new attacks which results in machine learning models producing low-quality results. In this study, a new dataset is proposed from widely used public cyber-attack tools being used by attackers in the real world. This dataset is also merged with two other commonly used datasets, CIC-IDS-2017 and CIC-DDOS-2019, to solve the problem of data imbalance in existing datasets.