An upper bound check issue in dsaVerify
function allows an attacker to construct signatures that can be successfully verified by any public key, thus leading to a signature forgery attack.
In dsaVerify
function, it checks whether the value of the signature is legal by calling function checkValue
, namely, whether r
and s
are both in the interval [1, q - 1]
. However, the second line of the checkValue
function wrongly checks the upper bound of the passed parameters, since the value of b.cmp(q)
can only be 0
, 1
and -1
, and it can never be greater than q
.
In this way, although the values of s
cannot be 0
, an attacker can achieve the same effect as zero by setting its value to q
, and then send (r, s) = (1, q)
to pass the verification of any public key.
All places in this project that involve DSA verification of user-input signatures will be affected by this vulnerability.
Since the temporary private fork was deleted, here's a webarchive of the PR discussion and diff pages: PR webarchive.zip
{ "nvd_published_at": "2023-10-26T15:15:09Z", "cwe_ids": [ "CWE-347" ], "severity": "HIGH", "github_reviewed": true, "github_reviewed_at": "2023-10-26T20:53:21Z" }