Due to design flaws, problems with implementations and improper network configuration, the Internet of Things devices become vulnerable in the network. They can be easily compromised and can also be attached to the Botnet network. IoT devices classification allows for strengthening of the overall network security through better VLAN planning and better firewall rule fine-tuning (e.g. per device class). In this paper only two classes of devices are considered: single-purpose devices (such as a bulb) and multi-purpose devices (such as mobile phone). Existing solutions do not provide the required accuracy within the given timeframe. We propose ML-based classification method based on supervised machine learning technology (Random Forest). With advanced packets flow analysis, our proposed approach demonstrates 94% of accuracy (7% better than the existing prior art). Additionally a very low False Positive rate is guaranteed for single-purpose IoT devices (e.g. a bulb must never be classified as a multi-purpose device).

Recently, adversarial attacks have drawn the community’s attention as an effective tool to degrade the accuracy of neural networks. However, their actual usage in the world is limited. The main reason is that real-world machine learning systems, such as content filters or face detectors, often consist of multiple neural networks, each performing an individual task. To attack such a system, adversarial example has to pass through many distinct networks at once, which is the major challenge addressed by this paper. In this paper, we investigate multitask adversarial attacks as a threat for real-world machine learning solutions. We provide a novel black-box adversarial attack, which significantly outperforms the current state-of-the-art methods, such as Fast Gradient Sign Attack (FGSM) and Basic Iterative Method (BIM, also known as Iterative-FGSM) in the multitask setting.