Temporal Adversarial Examples Attack Model Improves Network Intrusion Detection System Reliability

Strategic Insights -

Recent research by Dengpan Ye and colleagues from Wuhan University has proposed a novel recurrent neural network (RNN) adversarial attack model called Temporal Adversarial Examples Attack Model (TEAM). According to the study, the development of artificial intelligence has made neural networks crucial for network intrusion detection systems (NIDS). However, these networks are vulnerable to adversarial attacks, which can greatly impact the reliability of NIDS in real-world applications. To improve the reliability of NIDS, the researchers have proposed TEAM, a model that exploits the connection between adversarial examples and time steps in RNNs. The study shows that TEAM significantly improves the misjudgment rate of NIDS on both black and white boxes, achieving a misjudgment rate of over 97.65%.

Key Takeaways:

  • The research proposes a novel RNN adversarial attack model called Temporal Adversarial Examples Attack Model (TEAM) to improve the reliability of NIDS.
  • TEAM exploits the connection between adversarial examples and time steps in RNNs, making it effective against adversarial attacks.
  • Experimental results show that TEAM improves the misjudgment rate of NIDS on both black and white boxes to over 97.65%.
  • The study also finds that TEAM can significantly increase the misjudgment rate of NIDS for subsequent original examples by up to 95.57%.
  • The researchers conclude that existing solutions rarely consider adversarial attacks against RNNs with time steps, which is a critical oversight in the development of NIDS.
  • Financial support for this research came from the National Natural Science Foundation of China (NSFC).
  • The study has been peer-reviewed and published in IEEE Transactions on Network Science and Engineering.

Statistics:

  • The misjudgment rate of NIDS improved by over 97.65% with the use of TEAM. (Source: [Team: Temporal Adversarial Examples Attack Model Against Network Intrusion Detection System Applied To Rnn, IEEE Transactions on Network Science and Engineering, 2025])
  • The maximum increase in the misjudgment rate of the NIDS for subsequent original examples exceeded 95.57%. (Source: [Team: Temporal Adversarial Examples Attack Model Against Network Intrusion Detection System Applied To Rnn, IEEE Transactions on Network Science and Engineering, 2025])

Sources:

  • NewsRx. New Findings on Network Science and Engineering Described by Investigators at Wuhan University (Team: Temporal Adversarial Examples Attack Model Against Network Intrusion Detection System Applied To Rnn). Journal of Engineering. July 21, 2025; p 1780.
  • Team: Temporal Adversarial Examples Attack Model Against Network Intrusion Detection System Applied To Rnn. Ieee Transactions On Network Science and Engineering, 2025;12(4):3400-3415.