Improvising Encrypted Traffic Analysis Using Stacking Ensemble Model

Since the widespread adoption of network packet encryption, there has been a significant surge in the volume of encrypted data traversing networks. The majority of website traffic packets are now encrypted, necessitating decryption for readability. Attackers often exploit hosts using malware, making it challenging to identify packets containing fraudulent information without the necessary infrastructure in place. Presently, HTTPS serves as the default protocol for 70 \(\%\) of websites, as reported by W3Techs [1]. This prevalence underscores the importance of network traffic analysis, wherein security experts capture and meticulously scrutinize network packets. However, even with this analysis, it remains challenging for analysts to discern malicious encrypted data packets as they closely resemble regular ones. While decrypting packets presents a viable solution to this issue, it comes at a considerable cost and raises privacy concerns. The ability to detect malware within the network without decrypting packets offers analysts the opportunity to explore new features and techniques. This paper seeks to leverage machine learning techniques to identify harmful packets and detect malware in encrypted communication without the need for decryption.