A Framework to Identify Cyber Threats by Using Deep Learning Techniques

To identify some of the biggest cyber security threats the researchers are coming up with the automatic and efficient solutions. In order to improve cyber threat identification, the suggested strategy includes breaking down a substantial amount of gathered security occurrences into distinct event characteristics and applying an algorithm based on deep learning. The objective of this research is to develop an AI-SIEM system that integrates event profiling with various artificial neural network approaches for data preparation and also finding various cyber threats by using a dataset called SWAT and find the accuracy of different algorithms that are used so that this work can find the cyber threats in a more efficient way. This empirical results reveal substantial enhancements in classification performance across successive iterations of training, with notable metrics such as accuracy, F1-score, precision and recall consistently achieving remarkable numerical values across various attack types with deep learning methods of Auto encoder, PCA, DNN. For instance, it has been observed accuracy values ranging from 90.31% to 99.83%, F1 Score values from 73.62% to 95.35%, Precision values from 82.05% to 99.83% and Recall values from 67.25% to 92.51%. Overall, the findings help in cyber security in distributed and privacy-sensitive environments. This strengthens the robustness of modern cyber security frameworks.