Abstract
The various types of communication technologies and mobility features in Internet of Things (IoT) on the one hand enable fruitful and attractive applications, but on the other hand facilitates malware propagation, thereby raising new challenges on handling IoT-empowered malware for cyber security. Comparing with the malware propagation control scheme in traditional wireless networks where nodes can be directly repaired and secured, in IoT, compromised end devices are difficult to be patched. Alternatively, blocking malware via patching intermediate nodes turns out to be a more feasible and practical solution. Specifically, patching intermediate nodes can effectively prevent the proliferation of malware propagation by securing infrastructure links and limiting malware propagation to local device-to-device dissemination. This article proposes a novel traffic-aware patching scheme to select important intermediate nodes to patch, which applies to the IoT system with limited patching resources and response time constraint. Experiments on real-world trace datasets in IoT networks are conducted to demonstrate the advantage of the proposed traffic-aware patching scheme in alleviating malware propagation.
Introduction
The rapid expansion of the Internet of Things (IoT) has introduced significant security vulnerabilities. Most IoT end-devices, such as smart sensors or home automation components, possess minimal computing power and lack robust security features, making them ideal targets for malware like the Mirai botnet. Unlike traditional networks where infected nodes can be directly patched, IoT devices are often inaccessible or technically limited, precluding direct remediation.
Malware in these environments typically spreads through two primary channels: infrastructure links (e.g., cellular networks or the internet) and device-to-device (D2D) links (e.g., Bluetooth or Wi-Fi Direct). This dual-mode propagation allows malware to spread rapidly across vast distances via infrastructure while also moving locally between proximate devices.
System Model and Proposed Scheme
To address these challenges, this research proposes a shift in perspective: instead of patching billions of resource-constrained end-devices, security efforts should focus on "intermediate nodes"—the routers, access points (APs), and base stations (BSs) that serve as gateways for IoT traffic.
The core idea is to treat the problem as "link recovery." By securing these high-leverage intermediate nodes, operators can effectively sever the infrastructure links used by malware. This effectively confines the infection to local D2D dissemination, which is significantly slower and more manageable, creating a "virtual patch" for the infected devices behind the secured gateway.
Traffic-aware Patching Algorithm
The proposed Traffic-aware Patching scheme prioritizes nodes based on real-time traffic volume, which serves as a proxy for a node's importance in the network and its potential as a malware propagation hub. The algorithm operates in two distinct phases:
- Monitoring Phase: For a specified duration ($t_p$), the system monitors and collects traffic information for all intermediate nodes ($S_{AP}$).
- Patching Phase: Once the monitoring period ends, the system calculates an importance metric based on the observed traffic. Nodes are sorted in descending order of traffic volume, and the top $p\%$ of nodes are patched.
Finding the optimal monitoring time ($t_p$) is a critical trade-off. Patching too early results in insufficient data for accurate node prioritization, while waiting too long allows the malware to spread extensively before defenses are deployed.
Experimental Evaluation
The scheme was evaluated using a real-world trace dataset from a mobile social network consisting of 59 users and 1,751 Access Points. The simulation modeled realistic communication patterns and mobility.
Key findings include:
- Superior Performance: The traffic-aware approach significantly outperformed random patching strategies, especially when patching resources were limited (e.g., patching less than 30% of nodes).
- Optimal Timing: The experiments identified a "Goldilocks zone" for monitoring. In the specific simulation environment, a monitoring period of 40 seconds before patching yielded the maximum reduction in the fraction of compromised users.
- DDoS Mitigation: By prioritizing high-traffic nodes, the scheme is particularly effective against botnet-driven DDoS attacks that rely on high-volume traffic redirection.
Conclusion and Future Work
Securing the network infrastructure provides a practical and high-leverage solution for IoT security. While this method effectively bottlenecks infrastructure-based propagation, it does not stop local D2D spread. Future research will explore more sophisticated importance metrics, such as betweenness centrality, and the use of machine learning to adapt to zero-day threats. Additionally, developing notification mechanisms to alert users of local D2D risks remains a priority.
Summary updated