A Time Series Intrusion Detection Method Based on SSAE, TCN and Bi-LSTM

• DOI: 10.32604/cmc.2023.046607
• Publication Date: 2023-12-28T07:42:08Z
• Authors: Zhenxiang He; Xunxi Wang; Chunwei Li
• Keywords: Benchmark (surveying); Autoencoder; Adaptability

In the fast-evolving landscape of digital networks, incidence network intrusions has escalated alarmingly.Simultaneously, crucial role time series data in intrusion detection remains largely underappreciated, with most systems failing to capture time-bound nuances traffic.This leads compromised accuracy and overlooked temporal patterns.Addressing this gap, we introduce a novel SSAE-TCN-BiLSTM (STL) model that integrates analysis, significantly enhancing capabilities.Our approach reduces feature dimensionality Stacked Sparse Autoencoder (SSAE) extracts temporally relevant features through Temporal Convolutional Network (TCN) Bidirectional Long Short-term Memory (Bi-LSTM).By meticulously adjusting steps, underscore significance bolstering accuracy.On UNSW-NB15 dataset, our achieved an F1-score 99.49%, Accuracy 99.43%, Precision 99.38%, Recall 99.60%, inference 4.24 s.For CICDS2017 recorded 99.53%, 99.62%, 99.27%, 99.79%, 5.72 s.These findings not only confirm STL model's superior performance but also its operational efficiency, underpinning real-world cybersecurity scenarios where rapid response is paramount.Our contribution represents significant advance cybersecurity, proposing excels adaptability dynamic nature traffic, setting new benchmark for systems.