Skip to content

Back to the catalog

Associated-Data Correlations in Eevee: Deterministic Forgeries against Nonce-Misuse Security
eprint.iacr.orgpaper

Associated-Data Correlations in Eevee: Deterministic Forgeries against Nonce-Misuse Security

Cryptanalysis of the Eevee forkcipher-based AEAD modes Umbreon and Espeon, which claim online AE security under nonce reuse. Associated-data correlations give probability-one non-empty-message forgeries with three encryption queries and one verification query. The paper adds an affine treatment of partial final blocks (n-r+3 queries, or n-r+2 with adaptive choices). It also checks the published Umbreon attack with the official ForkSkinny primitive and reports a two-query attack accepted by the unmodified public C modes.

TL;DR Deterministic AD-correlation forgeries break nonce-misuse security of Eevee's Umbreon and Espeon AEAD modes, including a two-query attack on the public C code.
Document

What it is

Isobe and Rahman's ePrint 2026/2518 is a cryptanalysis of the Eevee forkcipher AEAD modes Umbreon and Espeon. Both claim online AE security under nonce reuse. The attacks are deterministic: probability-one forgeries on non-empty messages.

How it works

  • Two AD-only tags reveal a mask difference.
  • That difference lets an attacker translate a known ciphertext from one AD value to another, so the ciphertext verifies under the new AD.
  • The basic attack needs three encryption queries and one verification query.
  • Partial final blocks are handled with an affine construction. For a fixed final-fragment length r, it takes n-r+3 encryption queries.
  • If the forged length may be r or r-h (chosen 1 ≤ h < r < n) and the source plaintext is chosen adaptively, it takes n-r+2.
  • Both constructions satisfy the padding check.
  • They work for general message lengths. Espeon additionally needs at least one preceding complete block.

Preconditions / Affected

  • The target is Umbreon or Espeon deployed with nonce reuse, which is the setting where they claim security. The attacker needs encryption queries and one verification attempt.
  • The paper also analyzes a published short-message Espeon AD omission.

What to do

  • Treat the nonce-misuse (online AE) claims for Umbreon and Espeon as broken. Don't rely on them where nonce reuse is possible.
  • The paper gives no fix. Adding finalization is discussed only as an unproved countermeasure direction.
  • To reproduce, the authors report a two-query attack accepted by the unmodified public C modes. They also validated the published complete-block Umbreon attack using the official ForkSkinny primitive in a specification adapter.

Caveats

  • The basic attacks and their proof diagnosis overlap with independent, concurrent work by Mao et al. The authors' own contribution is the partial-block treatment.
  • The page is an abstract only. It gives no concrete parameters, test vectors or code location.
  • The page doesn't say whether the Eevee designers have responded.
  • It is a preprint and has not been peer reviewed.

Organisations named here

  • Espeon
  • nonce misuse
  • Umbreon

Technologies named here

  • Eevee Umbreon mode
  • Eevee Espeon mode
  • Unmodified public C implementation of Eevee modes
  • C
  • Eevee
  • ForkSkinny

Techniques named here

  • Associated-data correlation attack
  • Affine treatment of partial final message blocks
  • Analysis of the published short-message Espeon AD omission