Active defense mechanism for network security situation prediction based on transformer and TCN

With the continuous evolution of cyber-attacks, traditional network security posture prediction methods generally suffer from insufficient predictive accuracy and low computational efficiency in the face of large and dynamically changing data environments. This makes it difficult to meet today's demand for real-time defense and proactive response. In order to solve the above challenges, the research constructs a network security situation prediction model that integrates causal convolution and temporal convolution structures. The model introduces the causal convolution mechanism to enhanace the ability of the attention structure to model the causal and local mutation features in time series. Moreover, it uses the time convolution network to extract the local time features at multiple scales to achieve the synergistic modeling of global and local features. The results showed that the mean absolute error and root mean squared error of the research model for performing network security situation prediction were 0.0085 and 0.0143, respectively. Moreover, the mean time of network security situation prediction value with an average time was 213 ms, which was better than other models. The results demonstrate that the intelligent network security situation prediction model constructed in the study can quickly and accurately predict the network security situation, which enhances the intelligence level of network security defense. This can help network administrators deploy defense strategies in advance and reduce the damage caused by potential attacks.