| Title | GitHub Dropbear 2025.89 Improper Verification of Cryptographic Signature |
|---|
| Description | The specific issue is located in curve25519.c:
Line 484 directly uses s+32 for computation, with no validation for S >= L at any point.
According to the RFC 8032 §5.1.7 specification, an out-of-bounds S should be rejected. Without this check, we can generate an equivalent signature by constructing S' = S + L. My local testing confirms that Dropbear's verification process accepts both the original signature and its S+L variant. Furthermore, the paper "The Provable Security of Ed25519: Theory and Practice" by Jacqueline Brendel, Cas Cremers, Dennis Jackson, and Mang Zhao (https://ieeexplore.ieee.org/stampPDF/getPDF.jsp?arnumber=9519456) emphasizes that strictly rejecting such non-canonical signatures is necessary to maintain Strong Existential Unforgeability under Chosen Message Attacks (SUF-CMA).
Historically, there have been multiple signature malleability vulnerabilities in other projects (e.g., CVE-2020-36843, CVE-2024-45193). This inconsistency with the specification breaks signature uniqueness and could potentially impact higher-level security policies or auditing logic that rely on it. |
|---|
| Source | ⚠️ https://github.com/mkj/dropbear/issues/406 |
|---|
| User | pythok (UID 95793) |
|---|
| Submission | 02/23/2026 17:02 (1 month ago) |
|---|
| Moderation | 03/07/2026 10:07 (12 days later) |
|---|
| Status | Accepted |
|---|
| VulDB entry | 349652 [mkj Dropbear up to 2025.89 S Range Check src/curve25519.c unpackneg signature verification] |
|---|
| Points | 20 |
|---|