Key-Reuse Vulnerability of Phase-Keyed Fourier-Curve Modulation: Relation Leakage and Key-Refresh Cost on Coded Links

Reusing a phase key in harmonically coupled modulation converts short modular relations among the harmonic indices into estimable key characters, demonstrating that nominal key-space size and error rates are insufficient security evidence for keyed modulations with repeated waveform structure.

The gist

A modular relation lattice of the harmonic indices characterizes which key characters are exposed by data-cancelling mixed moments, and a non-data-aided relation-moment estimator turns this leakage into an attack that never enumerates the key space.

How it works: Relation Leakage Mechanism

The security argument based on key space size and wrong-key error rate fails when the waveform admits low-order statistics that cancel the data. For a phase-keyed Fourier-curve constellation, where each symbol occupies several complex tones whose phases are integer multiples of a data parameter, integer relations among the harmonic indices yield data-cancelling mixed moments of the received tones that expose key characters. Specifically, for consecutive harmonics, third-order moments recover the relative phases and a fourth-order moment completes the key up to cyclic relabeling whenever its coefficient is nonzero.

How it works: The Attack Methodology

The paper describes a non-data-aided relation-moment estimator that exploits this leakage. This attack evaluates the eavesdropper’s coded block error rate on an LDPC-coded link as a function of the number of symbols that share one key. The attack involves:

1. Using a fixed relation set S, consisting of exact-zero relations (S0) and modular relations with specific properties (S1).

2. Calculating the monomial moments, where the real coefficient Kc is derived from these moments.

3. Refining an estimate of the key phase using a Gauss–Newton approach on Pc in S0 to find a candidate key.

4. Using weights derived from S1 to refine the estimate further, and finally selecting a candidate based on segment-local parity-check counts and rounding rules, avoiding exhaustive enumeration of the key space.

How it works: Performance under Different Conditions

The study evaluates two harmonic families: consecutive harmonics, where third-order relations generate relations that recover relative phases, and odd harmonics, which require fourth-order relations to generate the lattice. The paper shows that for consecutive harmonics at (M, k) = (64, 16), the fourth-order relation completes L64(h), and if K4e16 is nonzero at evaluated noise levels like β = 0.3, the key is identifiable up to cyclic relabeling. Conversely, for odd harmonics at M=64, no nonzero c in L64(h) has a norm of one or two, and fourth-order relations generate the lattice.

How it works: Key-Refresh Cost Analysis

The paper ties the attack to key entropy by relating periodic independent refresh schedules to the secret entropy they consume. For an explicit finite product-grid protocol (2 128 keys), equallength refresh schedules that keep a 95% lower confidence bound of her block error rate above 0.9 consume at least 1.52 fresh key bits per information bit, which is 1.52 times the entropy rate of a one-time pad on the data. This cost is quantified by the metric ρK = H(K)/(LRc log2 M) fresh key bits per information bit, comparing it against a one-time pad on the information bits. The results show that for certain configurations, this consumption can be comparable to or greater than one-time pad encryption under selected operational rules.

How it works: Robustness and Comparison

The study tests robustness by introducing tangent artificial noise (AN) and comparing the consecutive family with odd harmonics. Tangent AN and the odd harmonic family without relations below order four raise Eve’s measured error rate at intermediate reuse lengths but do not remove the one-codeword vulnerability. The comparison between families reveals that while different in relation order, multiplicity, coefficient magnitudes, moment dependence, and geometry, their joint effect demonstrates that minimum relation order identifies which characters leak. Furthermore, the key-estimation step can be examined without decoding to predict the median key error using a Gaussian plug-in approximation based on the covariance of the relation-phase vector Xl.

How it works: Conclusion and Implications

The final analysis relates resource implications to how segment keys are supplied. Independent random refresh consumes H(K)/L new secret bits per symbol, and under the post-hoc 0.9 criterion, every tested schedule that frustrates the evaluated attack consumes at least as much as a one-time pad on the information bits for the 2 128-key grid. This demonstrates that nominal key-space size and error rate are insufficient security evidence for keyed modulations with repeated waveform structure, suggesting that cryptographic generators or physical-layer key generation must be assessed by their net secret-key rate after probing and reconciliation.

Here are the specific improvements that can be made to AI systems, derived from the insights in this scientific paper:

1. The development of a non-data-aided relation-moment estimator allows an eavesdropper to perform a key estimation attack without enumerating the entire key space. By understanding which high-order moments (specifically the fourth-order moment for consecutive harmonics, or specific relations for odd harmonics) reveal key characters, AI systems can be trained to recognize these leakage patterns in received signals.

2. AI systems can be designed to perform relation-moment attacks on intercepted communication streams. This means the system doesn't need a secret key to infer information about the key; it only needs knowledge of the modulation parameters (like harmonic set size, AN fraction, and noise level) and can estimate the key up to cyclic relabeling using techniques like Algorithm 1.

3. AI systems can be used for key-refresh cost analysis. By modeling how frequent key refreshes consume secret entropy relative to information bits (the ratio ρK), AI can optimize key management protocols in real-time. The system could recommend optimal refresh schedules that balance security against the computational/entropy cost of generating new keys, ensuring the required security margin (e.g., BLERE below 0.1) is maintained while minimizing entropy consumption compared to a one-time pad baseline.

4. AI systems can perform configuration-aware performance tuning. Since the paper shows that different families (consecutive vs. odd harmonics) and noise levels affect the attack's success, an AI system can dynamically adjust its detection strategy based on observed channel conditions (noise level, AN presence) to maximize its ability to detect key reuse exploitation or minimize error rates.

5. AI systems can improve detection robustness against non-ideal waveform structures. The paper demonstrates that tangent artificial noise (AN) and specific harmonic sets without low-order relations do not remove the one-codeword vulnerability but raise the measured error rate at intermediate reuse lengths. An AI system could be trained to specifically identify these robust configurations, distinguishing them from more vulnerable ones, thereby providing a more accurate assessment of link security.

6. AI systems can enhance key character identification. By analyzing the structure of modular relations (Proposition 1 and 2), an AI system can be trained to map observed signal statistics directly to specific key characters (e.g., identifying which phase offset corresponds to which key bit) using the derived relation lattice, enabling faster and more precise key recovery attempts than brute-force methods.