TL;DR
This paper presents a novel side-channel attack targeting hardware implementations of MAC-Keccak, a crucial component for data integrity and authentication. By analyzing power consumption during the first round of the Keccak permutation, the authors demonstrate that even hardware implementations are vulnerable to key recovery, offering a more efficient attack than previous methods.
What Problem Does This Paper Address?
The Keccak algorithm has been selected as the new SHA-3 standard, making its secure implementation critical. A Message Authentication Code (MAC) based on Keccak (MAC-Keccak) is essential for verifying data integrity and authenticity. While previous research has explored side-channel attacks on software implementations of MAC-Keccak, the security of hardware implementations, particularly on FPGAs, remained an open question. Hardware implementations often have different characteristics (e.g., parallel execution of operations) that might render existing software-focused attack models ineffective. This paper aims to bridge this gap by developing and demonstrating a practical side-channel attack specifically tailored for hardware MAC-Keccak.
Methods
The core of this paper is a novel side-channel attack that focuses on the first round output (R1) of the MAC-Keccak algorithm. Unlike previous attacks that targeted specific operations like the theta (θ) step, this method leverages the leakage from the entire first round. The authors propose new side-channel leakage models specifically designed for hardware implementations.
Their attack utilizes Correlation Power Analysis (CPA), a standard technique in side-channel analysis. A key innovation is the development of a more efficient attack strategy that targets entire rows of the state register rather than individual bits. This approach significantly reduces redundancy and improves the signal-to-noise ratio (SNR) of the leakage. The attack involves:
- Mapping the first round output (R1) to intermediate operations (θ, ρ, π, χ): By analyzing the relationships between R1 and the outputs of these operations, the authors can infer relationships between key bits and intermediate state bits.
- Targeting specific rows: Instead of analyzing each bit individually, the attack focuses on groups of five bits within a row of the state. This leverages the fact that these bits are processed together and their combined leakage is stronger.
- Key recovery through XOR relationships: The attack recovers not the key bits directly, but rather XOR combinations of key bits. By attacking multiple rows, these XOR relationships are combined to reconstruct the full secret key.
Datasets
The paper uses power traces from a hardware implementation of MAC-Keccak on an SASEBO-GII board. These traces were previously made available online. The authors do not introduce new benchmark datasets but rather utilize existing ones for their analysis.
Key Insights and Ideas
- Hardware MAC-Keccak is vulnerable: The study definitively shows that even hardware implementations of MAC-Keccak are susceptible to side-channel attacks, challenging the assumption that hardware inherently offers better security against such attacks.
- First round is a rich source of leakage: Focusing on the first round output (R1) proves to be an effective strategy, as it contains significant information about the secret key.
- Row-based attack is more efficient: Attacking entire rows of the state register, rather than individual bits, dramatically improves the efficiency and feasibility of the attack by increasing the SNR and reducing the number of traces required.
- MAC-Keccak is more resistant than other ciphers: Despite being vulnerable, the paper highlights that MAC-Keccak is more resistant to side-channel attacks compared to algorithms like AES and DES, especially when considering variable key lengths.
- Key length significantly impacts attack complexity: The paper explores how different key lengths affect the attack, demonstrating that longer keys increase the complexity for attackers, offering a built-in defense mechanism.
Key Results
- The proposed side-channel attack can efficiently recover secret key information from a hardware MAC-Keccak implementation.
- The attack requires approximately 35,000 traces to recover key bit relationships (one 3-bit XOR and four 2-bit XORs) for a single row with a success rate of 100%.
- The correlation between the Hamming weight of the first round output and power consumption is shown to be very strong, indicating a significant leakage.
- The attack method is demonstrated to be applicable to different initial states of the Keccak register, not just the all-zero state.
- The paper shows that increasing the key length significantly increases the difficulty of side-channel attacks.
Limitations and Future Work
- Initial state dependency: While the attack is shown to be applicable to different initial states, the specific leakage model and attack complexity might vary. The paper discusses the case where the register is initialized to random numbers, which significantly reduces leakage but doesn't eliminate it entirely.
- Key recovery is indirect: The attack primarily recovers XOR combinations of key bits, not the key bits directly. Recovering the full key requires combining results from multiple rows.
- Future work: The authors suggest investigating Keccak implementations that are more resilient to side-channel attacks and developing more efficient countermeasures. They also plan to explore attacks on implementations with key lengths greater than 320 bits.
Summary updated