TurboTLS: TLS connection establishment with 1 less round trip

Overview

• Presents a new approach called TurboTLS to establish TLS connections using one less round trip
• Sends the initial client-to-server and server-to-client flows of the TLS handshake over UDP instead of TCP
• Simultaneously carries out the three-way TCP handshake in the same flights
• Completes the final flight of the TLS handshake over the established TCP connection
• Avoids UDP fragmentation issues using request-based fragmentation
• Offers substantial latency improvements, especially on reliable connections

Plain English Explanation

The paper introduces a new technique called TurboTLS that can establish TLS connections more efficiently by reducing the number of round trips required. Typically, establishing a secure TLS connection involves a back-and-forth exchange between the client and server over a TCP connection.

In the TurboTLS approach, the initial steps of this TLS handshake are sent over a UDP connection instead of TCP. At the same time, the three-way TCP handshake to set up the underlying connection is also happening. Once the TCP connection is ready, the final step of the TLS handshake can be completed over that TCP link, and then the application data can be sent.

This allows the TLS connection to be established more quickly, effectively eliminating one round trip compared to the traditional approach. To avoid issues with UDP packet fragmentation, the system uses a clever technique where the client sends enough initial UDP requests to give the server enough space to fit its full response in a single packet.

Experiments show that this TurboTLS approach can provide substantial latency improvements, especially on reliable network connections where the round trip time savings are most impactful. Even on less reliable networks, the system has mechanisms to quickly fall back to the standard TCP-based TLS handshake if needed, ensuring adequate performance.

Technical Explanation

The core innovation in this paper is the TurboTLS technique, which modifies the standard TLS handshake process to leverage UDP in addition to TCP.

Traditionally, the TLS handshake occurs over a TCP connection, requiring multiple round trips between the client and server. In the TurboTLS approach, the initial client-to-server and server-to-client flows of the TLS handshake are sent over UDP rather than TCP. Simultaneously, the three-way TCP handshake to establish the underlying connection is also carried out in these same message flights.

Once the TCP connection is established, the client and server can complete the final flight of the TLS handshake over the TCP link and then use that connection for application data transfer. No changes are made to the contents of the TLS handshake protocol itself - only the delivery mechanism is modified.

To address potential issues with UDP packet fragmentation, the authors employ a request-based fragmentation technique. The client sends a series of UDP requests in advance, providing enough space for the server to fit its full response within a single response packet per request packet.

The authors also describe how clients can detect server support for TurboTLS without an additional round trip, by having the server advertise its capabilities in a DNS HTTPS resource record.

Experimental results demonstrate that TurboTLS can provide substantial latency improvements, especially on reliable network connections where the round trip time savings are most impactful. To ensure adequate performance on less reliable connections, the system employs lightweight packet ordering and buffering mechanisms, allowing the client to quickly fall back to the standard TCP-based TLS handshake if needed.

Critical Analysis

The TurboTLS approach presented in this paper offers a novel and promising solution for reducing the latency associated with establishing secure TLS connections. By leveraging UDP in addition to TCP, the authors are able to effectively eliminate one round trip from the standard TLS handshake process.

One key strength of this approach is its compatibility with existing TLS infrastructure - the paper does not require any changes to the TLS protocol itself, only the delivery mechanism. This should make it easier to adopt and deploy TurboTLS in real-world environments.

The authors also demonstrate thoughtful solutions to potential challenges, such as the UDP fragmentation issue. The request-based fragmentation technique seems well-designed to address this problem without adding significant complexity.

That said, the paper does not extensively explore the security implications of using UDP for the initial TLS handshake flows. While the authors note that no changes are made to the TLS protocol, it would be valuable to have a more thorough security analysis to understand any potential vulnerabilities or attack vectors introduced by this approach.

Additionally, the paper focuses primarily on latency improvements, but does not delve deeply into other performance metrics like throughput or CPU/memory usage. Further research could investigate the broader performance characteristics of TurboTLS to fully understand its tradeoffs and limitations.

Overall, the TurboTLS technique presented in this paper represents a promising step forward in optimizing TLS connection establishment. The authors have identified an important problem and developed an innovative solution. However, as with any new approach, ongoing research and real-world deployment will be necessary to fully evaluate its strengths, weaknesses, and long-term viability.

Conclusion

This paper introduces a novel technique called TurboTLS that can establish secure TLS connections using one less round trip than the traditional approach. By sending the initial TLS handshake flows over UDP while simultaneously setting up the underlying TCP connection, TurboTLS is able to reduce the latency of the connection establishment process.

The authors' experimental results demonstrate substantial performance improvements, especially on reliable network connections. The use of request-based fragmentation and other mechanisms to address potential UDP-related issues further bolsters the practicality of this approach.

While the paper does not explore all aspects of TurboTLS's performance and security implications in depth, it presents a compelling and innovative solution to an important problem in network communications. As web protocols continue to evolve and the demand for low-latency, secure connections grows, techniques like TurboTLS may play an increasingly important role in optimizing the underlying infrastructure.