Networks play a significant role in everyday life and cyber security has been a critical field of study.Rapid and efficient detection of network intrusion is regarded a basic framework in the cyber security field.While a large number of researches have concentrated on network intrusion detection over the previous years, developing an intrusion detection system with a strong detection accuracy and a comparatively low false alarm rate remains an issue.Several researches are focused on designing IDSs, which build on the techniques of Deep learning to overcome the above listed issues.This Paper used deep learning frameworks, including artificial neural networks and Generative Adversarial Networks, to outperform the earlier research.A structure of the generative adversarial networks, ID-GAN, is proposed in this paper to produce the adversarial attacks that mostly mislead and circumvent the method of intrusion detection.In comparison, it has used four databases in our research, related to intrusion detection systems to identify different conditions.Such databases include KDD CUP 99, CIDDS, NSL-KDD and CICIDS2017.In addition, 22 measurement criteria are used in each of the data sets to evaluate the efficiency of the system.Ultimately, at the end of this paper, comprehensive quantitative and rating tools are presented for interpretation of our findings.The results suggest a major change in our proposed method for detecting network attacks.