Optimal Packet Padding for IoT Traffic Obfuscation
Críston Pereira de Souza, Antonio Janael Pinheiro, Jeandro de M. Bezerra, Michele Nogueira · 2024
The Internet of Things (IoT) brings numerous benefits for society. However, it also allows the leakage of sensitive information (e.g., habits, behaviors, and health-related data), posing privacy risks. Solutions have tried to obfuscate network traffic and address this challenge. They typically involve padding and fragmentation techniques, that come with tradeoffs. This paper presents an optimal packet padding solution to define the most efficient number of bytes to insert into packets and achieve obfuscation. The approach is based on dynamic programming and relies on a discrete method, which exhibits the time complexity is Θ(mn2) and the memory complexity is Θ(mn). As the number of packets increases, memory consumption increases. Then, it introduces an approximate solution leveraging probability distributions to address scalability. Performance evaluation follows a trace-driven approach and employs metrics as accuracy, F1-score, and overhead. The proposal has obfuscated the identification of IoT devices by up to 50%, reducing overhead by up to 4 times compared to the traditional state-of-the-art methods.