added a comment explaining the process and tradeoffs

This commit is contained in:
James Brownlee
2024-01-02 20:02:28 -05:00
parent 4c872ec855
commit 46cbda08c4

View File

@@ -1,3 +1,34 @@
/*
This identity hiding process tests whether the rosenpass protocol is able to protect the identity of an initiator or responder.
The participants in the test are trusted initiators, trusted responders and compromised initiators and responders.
The test consists of two phases. In the first phase all of the participants can communicate with each other using the rosenpass protocol.
An attacker observes the first phase and is able to intercept and modify messages and choose participants to communicate with each other
In the second phase if the anonymity of an initiator is being tested then one of two trusted initiators is chosen.
The chosen initiator communicates directly with a trusted responder.
If an attacker can determine which initiator was chosen then the anonymity of the initiator has been compromised.
Otherwise the protocol has successfully protected the initiators identity.
If the anonymity of a responder is being tested then one of two trusted responders is chosen instead.
Then an initiator communicates directly with the chosen responder.
If an attacker can determine which responder was chosen then the anonymity of the responder is compromised.
Otherwise the protocol successfully protects the identity of a responder.
The Proverif code treats the public key as synonymous with identity.
In the above test when a responder or initiator is chosen what is actually chosen is the public/private key pair to use for communication.
Traditionally when a responder or initiator is chosen they would be chosen randomly.
The way Proverif makes a "choice" is by simulating multiple processes, one process per choice
Then the processes are compared and if an association between a public key and a process can be made the test fails.
As the choice is at least as bad as choosing the worst possible option the credibility of the test is maintained.
The drawback is that Proverif is only able to tell if the identity can be brute forced but misses any probabilistic associations.
As usual Proverif also assumes perfect encryption and in particular assumes encryption cannot be linked to identity.
One of the tradeoffs made here is that the choice function in Proverif is slow but this is in favour of being able to write more precise tests.
Another issue is the choice function does not work with queries so a test needs to be run for each set of assumptions.
In this case the test uses secure rng and a fresh secure biscuit key.
*/
#include "config.mpv"
#define CHAINING_KEY_EVENTS 1