A team of researchers from the Nanjing University of Science and Technology has introduced a powerful new cryptographic analysis framework that exposes critical vulnerabilities in Falcon, one of the premier post-quantum signature schemes standardized by the National Institute of Standards and Technology (NIST). Designated as Paper 2026/2888 on the Cryptology ePrint Archive and titled "FALCON-PARE: Full-Range Side-Channel Key Recovery of Falcon via Prior-Enhanced Evidence," the study was authored by Shuyi Chen, Yuejun Liu, Jingdian Ming, Wei Cheng, Yuhan Zhao, and Yongbin Zhou.
The research addresses long-standing limitations in side-channel cryptanalysis concerning post-quantum lattice-based cryptography. Specifically, it targets the number-theoretic transform (NTT) computations within Falcon, which constitute a primary and highly sensitive side-channel attack surface. While previous research has explored vulnerabilities in NTT operations, those methodologies largely relied on truncated coefficient ranges. Consequently, they failed to cover the full valid range of Falcon secret coefficients, severely limiting their real-world applicability for complete, end-to-end key recovery. The newly proposed FALCON-PARE framework successfully overcomes these barriers, offering a comprehensive, prior-enhanced evidence approach that achieves absolute recovery of the secret polynomials driving the signature scheme.
Falcon occupies a vital position in the modern cryptographic landscape. As organizations worldwide prepare for the advent of quantum computing—which threatens traditional public-key cryptographic algorithms like RSA and Elliptic Curve Cryptography—NIST has led an multi-year standardization effort to select robust post-quantum alternatives. Falcon, known for its remarkably compact signature sizes and efficient verification processes, emerged as a leading standard for applications requiring high throughput and strict bandwidth constraints. However, like many hardware- and software-implemented cryptographic schemes, its internal mathematical operations can emit subtle physical leaks through power consumption, electromagnetic radiation, or timing variations. These side-channel leakage vectors allow adversaries with physical access to the device to deduce internal states and cryptographic keys.
At the heart of the FALCON-PARE framework is a sophisticated synthesis of probabilistic modeling and hardware-level leakage extraction. The research team designed the system to combine Falcon’s intrinsic discrete-Gaussian prior with aggregated leakage data extracted from range conversion and Montgomery multiplication. By doing so, the framework successfully constructs high-fidelity input-layer soft evidence. This soft evidence is then systematically propagated through the complex structure of the NTT factor graph, allowing the analytical system to map out probabilities across a vast parameter space.
To refine these calculations and narrow down the possibilities, FALCON-PARE incorporates posterior entropy as a metric to select highly reliable coefficients. It then leverages inherent public-key relations to systematically resolve the remaining unknowns. Through this multi-staged analytical pipeline, the framework achieves the full, unabridged recovery of the critical secret polynomials $f$ and $g$ that underpin the security of the Falcon scheme.
To demonstrate the practical viability of their theoretical model, the researchers conducted systematic evaluations of FALCON-PARE on an STM32F407 microcontroller platform. The experiments were performed under distinct compiler optimization settings, specifically targeting -O0 and -O3 configurations. This rigorous testing environment allowed the team to validate the framework’s real-world practicality across multiple performance dimensions, including profiling-trace requirements, coefficient-wise accuracy, and the reduction of residual candidate spaces.
The empirical results underscore the potency of the FALCON-PARE approach. According to the study, the framework achieves single-trace full key recovery requiring just 40 profiling traces under the -O0 compiler setting, and 120 profiling traces under the -O3 setting. When benchmarked against existing soft-analytical side-channel attacks designed for NTT leakage, FALCON-PARE demonstrates dramatic performance improvements. Under identical comparison budgets, the new framework delivers absolute gains of 54.79 percent and 73.83 percent in coefficient-wise accuracy.
Furthermore, the framework drastically collapses the search space for attackers. Where previous methods left massive residual candidate spaces—approximating $2^4352$ and $2^5346$ under the respective compiler settings—FALCON-PARE successfully reduces the remaining candidate space to precisely one. This complete elimination of ambiguity highlights the severe nature of the vulnerability and emphasizes the necessity for robust countermeasures in practical deployments of post-quantum cryptographic standards.
The publication of Paper 2026/2888 serves as a significant milestone for the cryptographic community, illustrating that theoretical post-quantum security guarantees must be carefully matched with resilient implementation practices. As industries begin migrating toward NIST-approved algorithms like Falcon, findings from the Nanjing University of Science and Technology research team will undoubtedly inform future developments in side-channel protection, secure coding practices, and hardware hardening.

