Solana Networks Conducts Feasibility Study for Automating Labeling of Machine Learning Training Data

Solana Networks is pleased to announce that it recently completed a study with Defence R&D Canada for the efficient creation of labeled network traffic data.

Currently, there are very few publicly available labeled datasets which include different network traffic types and applications. There are even fewer available labeled datasets where the traffic is encrypted. This limitation hampers research and development efforts when using machine learning for network traffic analytics. Labeled or annotated network traffic is essential for training of machine learning based solutions. In addition, such data can serve as ground truth to be used for the evaluation and verification of machine learning systems and algorithms.

As part of this study, a high-quality labeled dataset of encrypted network traffic was created representing 10 traffic classes and 20 different applications. In addition, the team investigated the viability of algorithms to automate the process of labeled dataset creation. An in-depth study and rigorous evaluation of semi-supervised learning approaches such as co-training was carried out. The results of the study will be presented as part of the 15th International Conference on Network and Service Management, held in Halifax in October 2019.

For over 60 years, Defence R&D Canada, an agency of the Department of National Defence, has been providing scientific research and development of applied technology, and is a national leader in defence science and technology for the Department of National Defence and the safety and security communities.