A Swarm Intelligence-based Path Selection for Low-Power and Lossy Networks with the Presence of Packet Dropping Attacks

Abstract

Low-power and Lossy Networks (LLNs) have been the focus of growing interest in recent years due to their increased adoption of Internet of Things (IoT) and Machine-to-Machine (M2M) applications across various industries, such as smart homes, industrial automation, healthcare, and smart cities. An LLN is a network consisting of sensor nodes connected by lossy channels. The nodes are typically resource-constrained with limited battery power, memory size, and processing capabilities. And the channels are unstable, with relatively high packet loss and low packet delivery.

Due to the unique characteristics of LLNs that distinguish them from other network types, such as wired networks and ad hoc wireless networks, achieving both delivery reliability and routing efficiency is a challenging issue. Delivery reliability indicates the ability to successfully transmit data packets from a source node to a destination node with minimum loss and delay as possible. Routing efficiency, on the other hand, refers to the ability of a routing protocol to transmit data packets in a timely and resource-efficient manner. In other words, it measures how effectively a routing protocol can find the best path for data packet transmission, minimising energy consumption and reducing control overhead as possible. A routing solution designed for LLNs should operate within the constraints of lossy channels and limited node resources and effectively and efficiently achieve the routing function. This thesis attempts to address this challenging issue by designing and implementing a reliable path-finding solution for LLNs that takes into account all factors impacting packet delivery reliability and routing efficiency, including malicious threats, with the minimum cost possible. To this end, the thesis has made the following contributions.

Firstly, an in-depth study is carried out in the following areas of interest: LLNS requirements and environments, routing and path selection, and security threats affecting reliability. The purpose of the study is to (i) become familiar with the principle of LLNs, including the constraints and limitations of the network; (ii) identify reliability issues in the routing; and (iii) identify security threats, i.e. Packet Dropping Attacks (PDAs), affecting reliability. The intensive study was followed by a critical analysis of the state-of-art solutions in the context. This has led to the discovery that existing solutions are (i) mostly not designed for LLNs, (ii) mainly rely on a single metric for path selection, (iii) do not consider the effect of malicious attacks on reliability, and (iv) largely focus on simplified methods that can only make local optimum decisions.

Secondly, a novel Swarm Intelligence-based Path Selection (SIPaS) framework is presented. The SIPaS framework can achieve reliable path-finding in LLNs with the presence of Packet Dropping Attacks (PDAs). It contains two novel objective functions, an Ant Colony Objective Function (ACOF) and a Secured-Ant Colony Objective Function (S-ACOF). The SIPaS framework makes a global optimum decision for path selection based on multiple reliability-effecting factors. This selection decision is made in real-time in adaptation to the network topological changes caused by network fragmentation with as less energy consumption and control overhead as possible. The ACOF is designed to select the most reliable path between a pair of nodes in non-malicious LLNs, taking into account multiple routing metrics. The set of routing metrics used in ACOF measures (i) the quality of the path, (ii) the physical lengths of the links forming the path, and (iii) the remaining nodes' energy along the path. In contrast, the S-ACOF is designed to select the most reliable path between a pair of nodes in Malicious LLNs (MLLNs), taking into account an additional routing metric to measure the trustworthiness of the nodes.

Thirdly, the SIPaS framework ideas are implemented and evaluated using the Contiki Cooja simulator, a network simulator designed explicitly for LLNs. It allows (i) the simulation of different layers from the physical to the application layer and (ii) the emulation of the hardware of a set of sensor nodes. In addition, the Routing Protocol for Low-Power and Lossy Networks (RPL), pronounced as ripple, has been used as an underlying protocol for the framework. The simulation results showcase the exceptional performance of both ACOF and S-ACOF when compared to existing works in the domains of LLNs and MLLNs, respectively. ACOF has significantly improved Packet Delivery Ratios (PDRs), preserved energy, extended network lifetime, and reduced delay and control overhead. Its remarkable performance positions ACOF as a more reliable and effective solution for path selection in LLNs, surpassing the capabilities of related works. Similarly, S-ACOF stands out as a superior solution, particularly excelling in scenarios characterised by multiple malicious nodes, high node density, and heavy traffic loads. The simulation results indicate that S-ACOF outperforms other existing solutions in these challenging conditions. S-ACOF has the potential to address the critical challenges of MLLN routing, offering a more reliable and secure routing solution.