2016 OSDI OSDI 2016

Unobservable Communication over Fully Untrusted Infrastructure

Abstract

Keeping communication private has become increasingly important in an era of mass surveillance and state-sponsored attacks. While hiding the contents of a conversation has well-known solutions, hiding the associated metadata (participants, duration, etc.) remains a challenge, especially if one cannot trust ISPs or proxy servers. This paper describes a communication system called Pung that provably hides all content and metadata while withstanding global adversaries. Pung is a key-value store where clients deposit and retrieve messages without anyone— including Pung’s servers—learning of the existence of a conversation. Pung is based on private information retrieval, which we make more practical for our setting with new techniques. These include a private multi-retrieval scheme, an application of the power of two choices, and batch codes. These extensions allow Pung to handle 103× more users than prior systems with a similar threat model.

🌉 Interdisciplinary Bridge — Computer Science and Machine Learning
📈 Trend Setter — Cybersecurity
🧭 Keyword Pioneer — secure communication
🐝 Cross-Pollinator — Artificial Intelligence, Computer Science, Data Science & Analytics, Deep Learning, Knowledge & Reasoning, Machine Learning, Mathematics & Optimization, Natural Language Processing, Security & Privacy