Securing Ultra-dense Internet of Things Networks

In the next generation Ultra-dense (UD) Industrial Internet of Things (IoT) [called Ultra-Dense Networks (UDNs or UD-IoTs)], security is a non-ignorable argument/concern which is higher in UDNs than IoTs; we provide a rational, as below, on this common argument.;; A) The risk of zero-day attack detection in UDNs is multi-fold since the traditional [as well as Artificial Intelligence (AI) based] approaches for zero-day attacks detection are not effective in IoTs, let alone the UDNs; this is because the existing methods suffer poor performance issues. ;;B) Further, the problem is exaggerated organically when UDN scenarios are considered since thousands of mobile nodes dynamically join networks and therefore detecting unexpected traffic or suspicious scanning activity, in real-time, originating from a client or service is performance bounded, hence the performance is a key issue in the existing approaches. ;;One of the possible ways to mitigate such concerns is to re-investigate the routing protocols in the firewalls (data routing and policy design for secure routing to allow only necessary access/transactions) but they have their own limitations, including a lack of flexibility, overhead, congestion issues, non-adaptive to context change in context (concept evolution and concept drift), not elastic to high scalability as assumed in UDNs, and an inability to adapt to sudden changes in network traffic and scale; typically reiterating the performance related concerns discussed above. Further, in UDN scenarios, many gateways are deployed within small areas’ results in more signal interference compared to the classical situations results in more collision and delay leading to more dropped packets when traditional firewall approaches are used; further, high Packet Drop Rate (PDR) is a pressing issue both due to attack and non-attack scenarios which also impacts the data driven AI based real-time anomaly detection approaches in detecting zero-day cyber-attacks in Industrial Control Systems (ICS). ;;Overall, to prevent and protect UDNs from zero-day attacks we stand by with AI based anomaly detection approaches since they leverage statistical analyses to identify unusual behaviour of nodes, the approaches can be enhanced for mapping out the attack surfaces more carefully. Motivated from this, in the proposed project, we aim to design an architecture to prevent and protect zero-day attacks for UDN networks. ;;Inspired from this we aim to address the following research questions. ;;1. Firstly, we aim to investigate unique challenges in UDNs which limit the direct applications of the existing IoT's solutions in UDNs context. Different threat models, attack vectors, and stringent KPIs which are of-course different from IoTs, etc. are worth exploring.;;2. Secondly, we aim to investigate a novel approach by confining the problem statement into few application domains such as smart transportation, smart energy sector, etc. following this, we will evaluate and compare our solution with the state-of-the-art. ;;3. Finally, a Proof of Concept (PoC) using open-source real test bed will be given to show how the proposed approach would fit in with a real-life. We will use different metrics such as accuracy, confusion matrix, end-to-end delay, training time, and the overhead that may be generated by the proposal, etc. This is to demonstrate how the proposed approach would fit in with a real-life, this would capture the difficulties of deploying the solution in real network which will give a sense of realism to the solution. ;