TL;DR
This paper introduces the Offline Trusted Device and Proxy (OTDP) architecture, which enhances online security by using a proxy on an untrusted client device to manage communications with a separate, trusted device. This novel approach, enabled by a "TLS Switching" technique, aims to provide strong security assurances while maintaining transparency and usability for both users and applications.
What Problem Does This Paper Address?
The paper tackles the pervasive problem of compromised client devices, which pose a significant threat to secure user authentication and online transactions. Malware on client devices can intercept sensitive information like credentials, leading to fraudulent activities. While existing solutions often involve dedicated trusted devices, they can be cumbersome and lack transparency. The authors aim to bridge this gap by creating an architecture that leverages a trusted device for sensitive operations without disrupting the user experience or requiring modifications to existing server applications.
Methods
The core innovation is the Offline Trusted Device and Proxy (OTDP) architecture and its enabling mechanism, TLS Switching.
- OTDP Architecture: This architecture places a Proxy on the untrusted client device. This Proxy acts as an intermediary between the client application and server applications. Crucially, it also communicates with a Trusted App residing on an Offline Trusted Device (OTD). The OTD is responsible for handling sensitive computations and information.
- TLS Switching: This technique allows the Proxy to dynamically switch between two TLS communication channels:
- State T (Trusted): In this state, the Proxy is unable to read or write on the channel connecting the Trusted App to the server. This is achieved through TLS renegotiation, where the Trusted App renews session keys and initialization vectors, making them inaccessible to the Proxy due to forward and backward secrecy. This ensures sensitive data is exchanged only between the Trusted App and the server.
- State P (Proxy): In this state, the Proxy can efficiently handle communication between the client app and the server. The Trusted App passes keys and IVs to the Proxy, enabling it to read and write records through a standard TLS connection.
- Proxy Design: The Proxy is designed to be extensible and open-source, allowing users to inspect and modify its code. It supports various client interfaces (e.g., SOCKS, HTTP Proxy) and server interfaces, and uses a filter-based architecture for processing incoming data.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Transparency is Key: The architecture's success hinges on its transparency to both client and server applications. This means existing systems can adopt it without significant modifications, a major hurdle for many security solutions.
- Proxy as a "White Box": Making the Proxy open-source mitigates the trust issue. Users can verify its behavior, reducing reliance on the provider's trustworthiness.
- Untrusted Proxy, Trusted Device: The design cleverly separates concerns: the Proxy, running on an untrusted device, is not fully trusted, while the OTD is assumed to be secure. This pragmatic approach acknowledges the reality of client device compromise.
- TLS Renegotiation for State Management: The novel use of TLS renegotiation to dynamically switch between secure (State T) and efficient (State P) communication modes is a clever way to manage sensitive data flow without complex protocol redesign.
- Usability and Security Balance: The paper emphasizes that strong security is useless if it's not usable. OTDP aims to achieve this balance by minimizing user interaction for routine tasks while ensuring critical operations are handled securely.
Key Results
The paper presents a proof-of-concept implementation of the OTDP architecture and TLS Switching in Java. While specific quantitative results are not detailed in terms of performance benchmarks, the key findings are qualitative:
- Demonstrated Feasibility: The architecture has been implemented and demonstrated, notably using the OffPAD device.
- Potential Applications: The paper outlines several promising applications, including:
- Web Systems: Enhanced server app authentication (preventing phishing), user authentication (e.g., via biometrics), and secure modification of web pages to prevent insecure input.
- IoT Systems: Privacy watchdog for smart grid metering, secure authentication in e-hospitals, and securing communications in smart transportation.
- Security and Usability: The architecture is designed to offer strong security assurances against compromised client devices while aiming for a user-friendly experience.
Limitations and Future Work
- Limitations:
- Server App Vulnerability: The paper acknowledges that server apps might still be vulnerable to "lunchtime attacks" if they don't require user authentication and confirmation for sensitive actions, even when operating in State P.
- Client Device Integrity: While the Proxy is transparent, the client device can still arbitrarily modify received/sent information in State P, meaning the integrity and origin of information displayed or input on the client device cannot be fully trusted.
- Formal Verification: The security aspects have been considered but not formally verified.
- Future Work:
- Detailed User Experiments: Conducting more in-depth user studies to evaluate the architecture's usability.
- Formal Security Verification: Applying formal methods to rigorously prove the security properties of the architecture.
- Broader Authentication Methods: Integrating more widely used authentication protocols like OAuth.
- Fully Transparent Client Interface: Developing a client interface that doesn't require proxy configuration on client apps.
- Generalization: Extending the OTDP architecture for enterprise-specific needs and centralized credential management.
Summary updated