TL;DR
This paper introduces CoverUp, a system that "forces" website visitors to become involuntary participants in anonymous communication networks (ACNs). By leveraging basic browser functionality, CoverUp creates a large anonymity set, enhancing privacy for voluntary users and even hiding their intention to use the ACN, while addressing ethical and legal concerns.
What Problem Does This Paper Address?
Anonymous Communication Networks (ACNs) like Tor rely on a large number of participants to provide strong privacy guarantees. However, they face a "vicious cycle": a low number of users leads to weak anonymity, making the ACN unattractive, which in turn further reduces user participation. Even when anonymity sets are large, the mere act of using an ACN can raise suspicion. This paper tackles this challenge by proposing a novel approach: "forced" participation, aiming to bootstrap ACNs and provide plausible deniability for voluntary users.
Methods
CoverUp utilizes a web-based system that leverages JavaScript to make website visitors involuntary participants in an ACN. The core idea is to embed a JavaScript snippet in cooperating websites (entry servers) that instructs visitors' browsers to generate constant-rate cover traffic to a mix server. Two applications are built on this: an anonymous feed (uni-directional) and an anonymous chat (bi-directional).
- Entry Server: Hosts a webpage with an iframe request to the CoverUp server.
- CoverUp Server: Serves a JavaScript code snippet to the visitor's browser.
- JavaScript Code: Instructs the browser to generate constant-rate cover traffic to a mix server.
- Mix Server: Acts as a central point for mixing traffic.
- Voluntary Participants: Use a browser extension and an external application to actually retrieve feed content or send chat messages.
- Involuntary Participants: Only generate cover traffic via the JavaScript code, without installing any additional software or retrieving content.
The system is designed to make involuntary participants indistinguishable from voluntary ones to an attacker, thereby expanding the anonymity set.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- "Forced" Participation as a Solution: The paper argues that by "forcing" participation, ACNs can break the user participation cycle and achieve stronger anonymity.
- Indistinguishability is Key: The success of CoverUp hinges on making involuntary participants indistinguishable from voluntary ones to an attacker. This is crucial for both enhancing anonymity and providing plausible deniability.
- Plausible Deniability for Voluntary Users: By blending in with involuntary traffic, voluntary users can claim they were "forced" to participate, thus hiding their intention to use the ACN.
- Leveraging Existing Browser Functionality: CoverUp avoids requiring users to install complex software by using standard browser features like JavaScript and iframes.
- Ethical and Legal Considerations Addressed: The authors proactively discuss and address potential ethical concerns about "forcing" participation and explore the legal implications for websites and users.
Key Results
- Practical Performance: The implemented feed and chat applications demonstrated practical performance.
- Strong Privacy Guarantees: CoverUp provides strong privacy guarantees, particularly by creating a large anonymity set that includes all website visitors.
- Attacker Distinguishing Accuracy: With a year's worth of observations, an attacker could distinguish voluntary from involuntary users with an accuracy of only 56.3% in a realistic usage pattern, indicating a high level of privacy.
- Low Overhead for Entry Servers: The computational overhead for entry servers is minimal (just an iframe tag).
- Manageable Overhead for Involuntary Users: For non-mobile users, the traffic overhead (around 7.5MB per day) is considered negligible compared to streaming services.
Limitations and Future Work
- Trust Assumptions: The bi-directional channel (chat) requires trusting the CoverUp server for integrity checks of the JavaScript code. A malicious CoverUp server could potentially inject malicious code.
- Timing Leakage: While CoverUp aims to minimize timing leakage, the browser extension introduces some, which the authors acknowledge and discuss in their experimental evaluation.
- Potential for Suspicious Behavior: The paper notes that voluntary participants might spend more time on entry servers, which could be a potential privacy leak. They suggest alerting users or automatically closing tabs to mitigate this.
- Browser Profiling: The possibility of browser profiling to detect extensions is mentioned as a potential threat, though it's out of scope for this work.
- Future Work: The authors suggest extending CoverUp to a peer-to-peer variant for improved bandwidth and latency, making the mix server a full HTTP proxy for browsing arbitrary web pages, and exploring browser modifications for more efficient and less leaky extensions.
Summary updated