diff options
Diffstat (limited to 'directions/research_directions.tex')
| -rw-r--r-- | directions/research_directions.tex | 34 |
1 files changed, 31 insertions, 3 deletions
diff --git a/directions/research_directions.tex b/directions/research_directions.tex index c6222d9..1e2ba2f 100644 --- a/directions/research_directions.tex +++ b/directions/research_directions.tex @@ -598,6 +598,7 @@ $h$. % FIXME find and consistently apply a nice name for this handshake method \subsection{Alternative uses for interactive public channel binding} +\label{altuse} The channel binding method described above can be used in any scenario where a secure channel between two systems must be established where one party has a display of some sort and the other party has an input device of some sort. @@ -668,7 +669,6 @@ above, then copying the key through it. Compared to current common practice this transfer a key by simply reading out aloud the channel binding fingerprint. This reduces the problem of a digital out-of-band channel trusted for direct transfer of manipulation-sensitive key material to the problem of two users being sure whether they're actually talking to each other instead of an impostor. -% FIXME \section{Hardware implementation} \subsection{Hardware overview} @@ -691,14 +691,42 @@ Additionally, those operations are only invoked infrequently any time the device % separate power supplies, possible future filtering of power/gnd and comms link signals \subsection{Usability considerations} -% buzzer, leds etc. -\paragraph{Security-relevant UI components} +In many systems such as common TLS-based systems, overall system security heavily depends on implementation details such +as certificate checking and user interface details such as the precise structure of security warning messages and how +they can be overridden. The complexity of these components in practice often leads to insecure systems, such as a system +using TLS checking a certificate's internal validity but omitting checks on the certificate's chain of trust. A nice +property of the key estabilishment sytsem outlined in this paper is that it is both very simple, reducing surface for +errors and it tightly couples the critical channel binding step during key establishment to the overall system's user +interface. In a system using either keyboard or mouse-based interactive channel binding, an implementation that does not +perform the channel binding step correctly would simply not work. If the host does not display the correct fingerprint +the user cannot enter it and the device will not complete the binding step. If the device does not relay fingerprint +data correctly during pairing the host application would clearly indicate to the user things are amiss with the input +not matching the fingerprint. Since the channel fingerprint is computed in a cryptographically well-defined way based +on entropy contributed by both partners during pairing a implementer would not even be able to accidentially degrade +fingerprint security. + +The critical point from an UI perspective in this pairing scheme is that the host application must display correct +instructions to the user for them to complete pairing. In particular the host application must put emphasis on the user +actually checking whether the device raised an alarm before confirming pairing after fingerprint input. Even if it +didn't the user would notice the device not functioning, but an attacker might have gained unauthorized access in the +meantime. Likewise, the device needs a clearly understandable method of indicating pairing failure to the user. In +practice a loud buzzer and a few bright LEDs would likely suffice for this. \section{Evaluation} % FIXME \section{Future work} +The aspects outlined in section \ref{altuse} provide potential future research directions. The adaption of the system to +mouse input might be an interesting target for a user experience study, particularly in comparison with a purely +keyboard-based system. The SSH key exchange method would be an interesting target for a general-use systems +administration tool. Though we have done some basic security arguments in this paper, a more rigurous formalization +might be interresting for future use of this technology. We have soundly argued about the user experience benefits of +our method, but we have not performed any field experiments to back up these arguments. Future research might analyze +the practical security a system as outlined in this paper yields under real-world conditions. The various trade-offs of +e.g. keyboard vs. mouse input, fingerprint lenght and design details of the pairing UI might be analyzed with respect to +practical usability and security. + \section{Conclusion} % FIXME |