AI-Driven Security Paradigms: Elevating Cloud Protection with Machine Learning

In the literature, some studies have explored classifying network traffic using Long Short-Term Memory (LSTM) networks to enhance cloud security. We analyzed a dataset—BCCC—that includes various types of network traffic: Benign, Benign-Email-Receive, Benign-Email-Send, Benign-FTP, Benign-SSH, Benign-Systemic, Benign-Telnet, and Benign-Web_Browsing_HTTP-S. Key features examined include fwd_ack_flag_percentage_in_fwd_packets, fwd_ack_flag_percentage_in_total, min_fwd_header_bytes_delta_len, and handshake_duration. The model performed well in detecting the Benign class, but some classes with fewer samples, such as Benign-FTP and Benign-Email-Receive, require improved precision and recall due to class imbalance. Overall, the model’s performance in classifying network traffic is strong. This research outlines strategies for addressing class imbalance and refining feature engineering. It provides a foundation for further, more detailed investigations into AI approaches for network traffic classification, highlighting the importance of sample balancing to achieve high accuracy.