In today’s digital age, network security is critical as already billions of computers around the world are connected with each other over networks. It is therefore important to be able to detect anomalous network traffic, e.g. related to attacks like denial of service (DoS) and probing. Network anomaly detection (NAD) is an attempt to automatically detect potentially anomalous behavior that differs in some way by observing traffic data over time. However, network traffic data are real-world data compounded by properties such as large scale, noisy label, and class imbalance, making it a challenge for deep learning algorithms. For example, anomalies rarely occur and the majority is normal data (i.e. anomalies only typically occur 0.001-1% of the time), and learning from imbalanced data is still an open challenge.
The number of connected computers will increase exponentially as new technologies such as 5G, 6G (a SAL lighthouse topic), and beyond emerge. Bernhard Lehner is working in the Embedded AI team on anomaly detection (among other deep learning related topics) with a focus on real-world data and the challenges that are related to the physical world, which are too often ignored in the academic world.
About the challenge
The ZYELL-NCTU NAD challenge is an event of 2021 IEEE International Conference on Acoustics, Speech and Signal Processing, the biggest conference on this topic, therefore gains a lot of attention from both academia and industry. It was an opportunity to learn something new and compete with international research teams during a collaboration among Dynatrace Research, Silicon Austria Labs, Linz Institute of Technology AI Lab, and the Institute of Computational Perception, at the Johannes Kepler University of Linz. We managed to end up on the official leaderboard with a rank 8 out of what seems to have been at least 180 competitors.
6G research at SAL
SAL in Linz has been researching the next generation of mobile communication, 6G, for over 1,5 years. 6G enables the transmission of large amounts of data in real time, so it is characterized by low latency times and high reliability. In the “factory of the future”, not only humans and machines are networked, but also the systems. Devices can perceive their environment via sensor data, make decisions themselves and communicate with other devices or the cloud - with the result to make processes highly efficient. Our researchers work here in the fields of millimeter-wave RF technology, embedded artificial intelligence for communication, radar and sensing, and wireless communication for industrial applications. With this research focus, SAL aims to become a leading European center for 6G research and technology for industrial applications.