TL;DR
This paper introduces the Secure and Private Socket Layer (SPSL) protocol, an extension of SSL/TLS, designed to protect user privacy in subscription-based remote services. SPSL allows users to prove they are authorized without revealing their specific identity, addressing a critical privacy gap in current certificate-based access control systems.
What Problem Does This Paper Address?
In subscription-based services (like online newspapers or databases), service providers need to ensure only paying customers access their content. A common method is using X.509 certificates and the SSL/TLS protocol for authentication. While this effectively prevents unauthorized access, it poses a significant privacy risk to users. The authors highlight three main threats:
- Eavesdropping: Network traffic can be monitored, allowing an eavesdropper to build a profile of users and their access frequencies.
- Service Manager Snooping: The service manager, while verifying certificates for access control, can collect detailed information about what specific content a user accesses, even if this information is not essential for managing subscriptions. This can lead to detailed user profiling beyond what's necessary.
- Certificate Extension Disclosure: X.509 certificates can contain various extensions (e.g., medical conditions, job status). When a certificate is presented, all extensions are typically revealed, even if only a subset is relevant to the service being accessed, leading to unintended privacy breaches.
The core problem is the conflict between the service manager's need to verify authorization and the user's right to privacy, especially concerning the disclosure of personal information embedded in certificates.
Methods
The paper proposes two main innovations:
- Crypto Certificates: A modification to the X.509 certificate structure that allows for selective disclosure of extensions. This means a user can present a certificate where certain sensitive fields are hidden, and only reveal them when necessary and verifiable. This is achieved by using secure hash functions and random paddings for sensitive fields, allowing the certificate to be verified without revealing the actual data.
- Secure and Private Socket Layer (SPSL) Protocol: An extension of the SSL/TLS handshake protocol. The key idea is to allow a user to prove they possess one private key from a set of qualified users' public keys, without revealing which specific key they possess. This is based on a "basic anonymous identification protocol" that uses RSA keys. SPSL introduces different "anonymity levels" (NULL, WEAK, MEDIUM, STRONG) to control the degree of privacy offered.
- NULL: Equivalent to standard SSL/TLS.
- WEAK: Protects against eavesdropping by encrypting the client certificate.
- MEDIUM: The server presents a list of qualified certificates (a "mask"), and the client proves knowledge of a private key corresponding to one of them.
- STRONG: Similar to MEDIUM, but the mask is encrypted.
The authors also developed an SPSL API and implemented a private navigation system, including an SPSL-enabled Apache web server module and a proxy for Netscape Navigator, to demonstrate the viability of their approach.
Datasets
Not discussed in the paper. The paper focuses on protocol design and implementation rather than training on specific datasets.
Key Insights and Ideas
- Privacy is more than just anonymity: The paper argues that true privacy in this context involves not just hiding who you are, but also controlling what information about you is revealed, even when you are authorized.
- X.509 certificates are a privacy liability: The standard way certificates are used inherently leaks more information than is necessary for access control.
- Selective disclosure is key: The concept of "crypto certificates" allows for fine-grained control over what information is shared, making certificates more privacy-preserving.
- SPSL offers a spectrum of privacy: The different anonymity levels in SPSL allow users and services to choose a balance between privacy and functionality.
- Compatibility is crucial: The SPSL protocol is designed to interoperate with existing SSL/TLS infrastructure, making adoption easier.
Key Results
The paper presents a functional implementation of the SPSL protocol and a private navigation system. Experimental results show that while SPSL introduces some overhead compared to standard SSL/TLS, it is manageable, especially when using identifiers for certificate lists. The performance impact grows with the size of the qualified user set, but the authors demonstrate that by using subsets of certificates, SPSL remains viable even for large-scale scenarios. For instance, with a mask of 10,000 users, using an identifier for the certificate list resulted in a slowdown of only 7.4 seconds, which is considered acceptable.
Limitations and Future Work
- Network-level privacy: The authors acknowledge that SPSL itself does not hide network-level information (like IP addresses). They suggest combining SPSL with systems like Onion Routing or CROWDS for full untraceability.
- Server adoption: A significant hurdle is convincing servers to adopt SPSL, as it requires changes to their infrastructure and a willingness to support anonymous identification.
- Crypto certificate integration: While the concept of crypto certificates is proposed, their full integration and standardization within existing PKI frameworks is a future direction.
- Per-access billing: The current SPSL protocol is designed for subscription-based services and is not directly applicable to per-access billing models, which would require different cryptographic primitives like unduplicatable anonymous tokens.
Summary updated