Trace files for ML-based IoT intrusion detection
The ITEA project STACK (Smart, Attack-resistant IoT Networks) will enable Internet of Things (IoT) applications with a high quality of service even under non-benign circumstances. Goals include more robust IoT communication, attack detection/mitigation through performance and interference monitoring and algorithms leveraging a tight integration with a smart edge.
Lack of reliability and security of IoT networks
Many IoT networks lack guaranteed reliability, latency and security – a concern given the rise in cyberattacks. IoT mesh networks of embedded devices are especially vulnerable due to their wireless communication and relatively low output power; even so, they increasingly influence our safety and livelihood in domains like autonomous driving and healthcare. As resource constraints prevent devices from running sophisticated defences, the challenge is to ensure that IoT networks can maintain functionality in difficult situations such as attacks and cross-technology interference.
Intrusion detection systems (IDS) play an integral part in the defence against attacks. To be effective in various attack scenarios, IDS typically rely on machine learning-based algorithms. These algorithms require a lot of training data. In many domains, such data traces exist. In the area of low-power IoT mesh networks such data traces do not exist.
To this end, Multi-Trace has been designed in the STACK project. Multi-Trace extends Cooja, a widely used simulation tool for IoT mesh networks to generate data traces. These traces support the training of machine-learning algorithms for intrusion detection in low-power, resource-limited IoT networks. Multi-Trace’s trace generation facility includes logs at different levels. In addition, there are scripts to define simulation scenarios. Since all logs stem from the same simulation, they inherently share a global timestamp. This enables users to merge and combine different logs according to their needs and finally get data to train the machine learning algorithms to defend against attacks on IoT networks. Multi-Trace is available at STACK’s github: https://github.com/STACK-ITEA-Project