Abstract: We present practical attacks on IAR-CTR and IAR-CFB, AE modes proposed by Hwang and Gope to offer both privacy and authenticity in a single pass. Then we use the idea of iCOFB, which is originally defined with a tweakable random function, to fix the IAR-CTR mode using the tweakable block cipher, and propose the TIAR-CTR mode, which is a one-pass AEAD mode with provable security. However, the same fixed method does not work for the IAR-CFB mode.
Introduction
In the world of cryptography, efficiency is king—especially for real-time systems. Authenticated Encryption (AE) modes that can provide both privacy and integrity in a single "pass" over the data are highly sought after. However, as demonstrated by Zhu et al., achieving this efficiency without sacrificing security is a delicate balancing act. This paper examines the vulnerabilities of IAR-CTR and IAR-CFB and proposes a robust alternative.
The Problem: Vulnerabilities in One-Pass AE
Authenticated Encryption (AE) ensures that data remains both secret (privacy) and untampered with (authenticity). While traditional methods like AES-GCM are standard, researchers seek "one-pass" schemes that are faster and require less memory for real-time applications.
In 2015 and 2016, Hwang and Gope proposed IAR-CTR and IAR-CFB—modes designed to add authenticity to standard stream ciphers using a method called "intentional delay." The idea was to XOR plaintext blocks into the encryption process with a delay, eventually producing a tag. This paper addresses a fundamental question: Are these efficient modes actually secure? The authors prove they are not, showing that an attacker can forge valid messages with 100% probability using just a single query.
Cryptanalysis and the "Intentional Delay" Trap
The original IAR modes attempted to create a dependency between all blocks and the final tag. However, the authors found that this dependency was too "shallow." Because the randomization didn't change per block, the XOR sum of the ciphertext and plaintext ($C \oplus M$) effectively leaked enough information to bypass the integrity check.
The authors developed a forgery attack that exploits the linear relationship between the ciphertext, plaintext, and the resulting authentication tag. By intercepting a single (IV, C, T) tuple, an adversary can manipulate the ciphertext and calculate a new, valid tag without knowing the secret key.
Redesign: The TIAR-CTR Mode
To fix these critical flaws, the authors moved away from standard block ciphers (like AES) and adopted Tweakable Block Ciphers (TBC). In a TBC, the encryption function takes an extra input called a "tweak."
The "aha moment" in the fix is using the block position and a nonce as a tweak. By making the tweak unique for every single block ($t = N || i || b$), the algorithm ensures that even if an attacker knows the input to one block, they cannot predict the output of the next because every block uses a functionally independent permutation. This redesign, named TIAR-CTR, achieves provable security and supports associated data (AD).
Key Results
The findings are stark:
- Total Break of IAR-CTR/CFB: The authors demonstrate an existential forgery attack requiring only one encryption query to succeed with probability 1. This effectively breaks the authenticity claims of the original schemes.
- Privacy Collapse: The forgery attack can be extended to a Chosen Ciphertext Attack (CCA), meaning an attacker can not only forge messages but also potentially recover secret information.
- TIAR-CTR Security: The proposed replacement supports Authenticated Encryption with Associated Data (AEAD) and nonces. It is proven secure in the standard model, with a security bound related to the number of blocks processed.
Limitations and Future Work
Interestingly, the authors found that while their TBC-based fix worked for the Counter (CTR) mode, it failed for the Cipher Feedback (CFB) mode. In the CFB version, the tag became independent of the ciphertext, allowing an attacker to change the message entirely while keeping the same tag. This highlights that "one size does not fit all" when fixing cryptographic modes.
The most significant limitation noted is that the proposed fixing method does not work for IAR-CFB. Future work may involve looking for a more efficient way to secure CFB-based stream ciphers or exploring how TIAR-CTR performs in hardware-constrained environments compared to established standards like AES-GCM or OCB3.
Summary updated