CSTNET

The increasing prevalence of encrypted traffic inmodern networks poses significant challenges for network security,particularly in detecting and classifying malicious activitiesand application signatures. To overcome this issue, deep learninghas turned out to be a promising candidate owing to its abilityto learn complex data patterns. In this work, we present adeep learning-based novel and robust framework for encryptedtraffic analysis (ETA) which leverages the power of BidirectionalEncoder Representations from Transformers (BERT) and LongShort-Term Memory (LSTM) networks. Our proposed frameworkleverages the capability of LSTM to capture long-termdependencies in sequential data for modeling the temporal patternsof network packets, while BERT enhances this by providingan understanding of the bidirectional context within packetsequences. Hence, this approach of ETA relies on LSTM forenabling effective detection of anomalies and prediction of futurepacket where BERT provides a deeper contextual understandingof the traffic flow. Publicly available dataset ISCXVPN2016and CSTNET are used to test our proposed framework whichoutperformed the existing works by yielding an accuracy rate(AC) of 99.65%, precision (PR) of 99.53% and recall (RC) of99.28%. The proposed framework serves to efficiently detectOver-the-Top (OTT) application signatures within encryptedtraffic streams, ensuring comprehensive network monitoring andenhanced security measures without compromising the integrityof packets.