Executive Summary : | The increasing number of IoT devices is making managing the entire infrastructure and network increasingly challenging. With over 50 billion devices, it is necessary to classify them for various purposes such as configuration, monitoring, controlling, and analysis. However, the increasing number of attacks, such as DDos, sYN-flood, smurf, POD, and ARP spoofing, can lead to unavailability of legitimate services, performance reduction, latency, and average device life. To address these challenges, machine intelligence techniques are needed to classify IoT devices using network traffic characteristics. This solution could be applied over cloud networks to centralize secured systems, but an edge-based solution can help filter out malicious data and devices not performing regular behavior. This security patch can be used in various domains such as healthcare, transportation, smart cities, smart grids, wearable devices, industrial internet, and smart homes. The project proposal aims to construct an interoperable IoT dataset by deploying a physical testbed for Indian scenarios/networks, pre-process and feature engineering over captured traffic logs using privacy-aware deep learning models, and create a resilient framework for classification and detection of attacks in the network system. The proposed dataset will show the uses and patterns of IoT devices for Indian users, which will differ from the existing dataset. The framework will be designed and fine-tuned based on existing traffic in various industries and homes, providing resilience without privacy leakage. Federated learning-based techniques will be used to equip the privacy-awareness in the framework. |