I think the issue here is the ambiguity in the behavior. If App A knows App B’s behavior fully and with no ambiguity and App B knows how App A will respond fully and with no ambiguity, then it’s unlikely to be a problem.
The issue is that with the ambiguity, App A cannot fully know/anticipate App B’s behavior and/or vice versa. These ambiguities might be harmless within App A and within App B, but when they are exchanging information it becomes harmful. Indeed, this sort of vulnerability is more dangerous when happening between independent components than if it was just an internal issue (although this could also cause problems with internal communication when multiple engineers have different interpretations and later maintainers might not know the implicit choices), since it is highly likely that the designers of independent components will have different implicit assumptions. This line from the original paper I think supports my interpretation: “The disagreements on the interpretation of the OpenPGP standard may raise doubts on the interoperability between the libraries… While in a basic scenario these three libraries can, to the best of our knowledge, in fact interoperate securely, we shall now see that some choices made by Crypto++ and gcrypt prove to be fatal in a broader context.” (pg. 5) So at least part of the issue here is a failure of requirements/documentation, as much as it is a matter of handling potentially dangerous input (although that is also relevant for defense-in-depth). Is there a CWE for ambiguity in security protocols between multiple parties? With regards, John Thomas From: Kurt Seifried <k...@seifried.org> Sent: Thursday, September 23, 2021 11:20 PM To: noloa...@gmail.com Cc: cwe-research-l...@lists.mitre.org Subject: Re: Cross-configuration attacks I assume by CVE you meant CWE, and no there isn't a CWE for "intersection" or "mismatch" attacks. I don't like the term cross-configuration unless it's actually applied to issues that are created by configuration issues, my concern would be technically any intersection vulnerability can be classed as a config issue because you could disable most things somehow/somwhere. Perhaps we need CWE to not just cover weaknesses but normal behaviours so we can better describe "normal behaviour A + normal behavior B = weakness [described if not specific term exists). Do we have a list of CVE "intersection" vulns to look at as a data set to see what is causing these? E.g. configs? badly written specifications that result in different interpretations? One good keyword is "conjunction" but also a lot of false positives: https://nvd.nist.gov/vuln/search/results?form_type=Basic&results_type=overview&query=conjunction&search_type=all&isCpeNameSearch=false On Thu, Sep 23, 2021 at 8:16 PM Jeffrey Walton <noloa...@gmail.com<mailto:noloa...@gmail.com>> wrote: Hi Everyone, This made my radar recently: https://eprint.iacr.org/2021/923.pdf. The interesting thing about the attack is, App A is considered secure in isolation, and App B is considered secure in isolation, but when interacting App A and B produce an insecure result. We've seen bad interactions among components within the same app before, like incorrectly combining authentication and encryption. But in this case it is not the same app. Rather, the vulnerability is a product of two distinct apps using slightly different implementation details sharing data. I'm wondering if there's a CVE to cover the scenario. Looking through existing CVEs I don't see one that jumps out at me. ----- Here's from the abstract of the paper: ... ElGamal encryption has been used in many different contexts, chiefly among them by the OpenPGP standard. Despite its simplicity, or perhaps because of it, in reality there is a large degree of ambiguity on several key aspects of the cipher. Each library in the OpenPGP ecosystem seems to have implemented a slightly different “flavour” of ElGamal encryption. While –taken in isolation– each implementation may be secure, we reveal that in the interoperable world of OpenPGP, unforeseen cross-configuration attacks become possible. Concretely, we propose different such attacks and show their practical efficacy by recovering plaintexts and even secret keys. -- Kurt Seifried (He/Him) k...@seifried.org<mailto:k...@seifried.org>