Abstract
Galois Field arithmetic blocks are the key components in many security applications, such as Elliptic Curve Cryptography (ECC) and the S-Boxes of the Advanced Encryption Standard (AES) cipher. This paper introduces a novel hardware intellectual property (IP) protection technique by obfuscating arithmetic functions over Galois Field (GF), specifically, focusing on obfuscation of GF multiplication that underpins complex GF arithmetic and elliptic curve point arithmetic functions. Obfuscating GF multiplication circuits is important because the choice of irreducible polynomials in GF multiplication has the great impact on the performance of the hardware designs, and because the significant effort is spent on finding an optimum irreducible polynomial for a given field, which can provide one company a competitive advantage over another.
Introduction
In the competitive landscape of integrated circuit (IC) design, the specific implementation details of cryptographic hardware are often highly valuable intellectual property. For hardware implementing Galois Field ($GF(2^m)$) arithmetic, performance is heavily dependent on the choice of an irreducible polynomial ($P(x)$). Designers invest significant engineering resources to identify optimal polynomials that minimize area and delay for specific applications.
However, once a chip is deployed, reverse engineering (RE) techniques can extract gate-level netlists to identify the specific $P(x)$ used, allowing competitors to replicate optimized designs. While generic logic locking exists, it often introduces significant overhead or remains vulnerable to SAT-based attacks. This research proposes a specialized "Algorithmic Obfuscation" technique that leverages the mathematical structure of GF arithmetic to provide robust, low-overhead protection.
Design Methodology
The proposed approach focuses on making a single circuit appear to support multiple valid irreducible polynomials, while only one is functionally correct. The design flow involves initializing a multiplier with a primary polynomial and then generating additional logic to embed alternative functions, creating an obfuscated multiplier that presents a challenge to attackers.
Algorithmic Obfuscation
The core insight of this work is that the majority of a GF multiplier—the partial product generation—is identical regardless of the polynomial used. The differences are concentrated entirely in the reduction logic, where the choice of $P(x)$ manifests as a specific network of XOR gates.
To protect this logic efficiently, the researchers employed several key techniques:
- Camouflaged Standard Cells: Physically indistinguishable cells (using doping or dummy contacts) act as "switches" to select the active polynomial by acting as constant '0' or '1'.
- Logic Sharing: An algorithm named
structureGenidentifies the maximum common logic between different reduction structures, ensuring that only the differing bits are obfuscated. - Iterative Obfuscation: The method allows $n+1$ different functions to be embedded. While the designer knows the true function, an attacker sees multiple valid-looking options.
- Optimization: Early constant propagation and obfuscation term reduction are used to merge equivalent logic paths, minimizing the hardware footprint.
Evaluation and Results
The technique was evaluated using standard irreducible polynomials for degrees $m=8, 64, 128, 256$ and a 14nm technology library. The results demonstrate high efficiency and resilience:
- Performance Overhead: On average, adding an extra obfuscated function costs only 1.8% in area and delay. For a 256-bit multiplier (standard for ECC), obfuscating 16 different functions resulted in only a 20% area overhead.
- Security Resilience: State-of-the-art SAT-based attack tools failed to de-obfuscate a 16-bit multiplier within a 12-hour limit. Binary Decision Diagram (BDD) attacks also failed due to memory exhaustion (exceeding 16GB).
- Complexity: The researchers found that the mathematical complexity of GF arithmetic leads to a "memory explosion" for automated attack tools as bit-width increases, making the obfuscation increasingly robust for larger fields.
Conclusion
This algorithmic obfuscation technique provides a highly efficient means of protecting GF multiplier IP. By focusing on the reduction logic and maximizing logic sharing, the method achieves significant security with minimal performance impact. Future work may involve using machine learning to optimize the selection and ordering of polynomials to further reduce overhead. While currently optimized for multipliers, the principles of this research highlight a path forward for mathematically-grounded hardware security.
Summary updated