Fuzzy Covert Time-Lock Puzzles via Sequential Noise Stripping
Classical Time-Lock Puzzles (TLPs) enforce a strict sequential computational delay $\mathcal{T}$ prior to secret recovery, yet they exhibit three structural limitations in adversarial and censorship-prone peer-to-peer (P2P) networks: (i) public decodability, whereby any party completing $\mathcal{T}$ sequential steps recovers the plaintext; (ii) exact-match rigidity, where restricting post-delay access via symmetric encryption induces combinatorial key-management overhead, precludes fuzzy (error-tolerant) attribute matching, and leaks cleartext key identifiers; and (iii) algebraic wire visibility, as structured puzzle parameters (e.g., RSA moduli or group elements) yield deterministic $\mathcal{O}(1)$ distinguishers for passive Deep Packet Inspection (DPI). In this paper, we introduce the Fuzzy Covert Time-Lock Puzzle (FCTLP), a hybrid cryptographic primitive unifying sequential time-delay release with fuzzy attribute-based access control and zero-metadata steganographic transport. Rather than attempting to derive sequentiality from parallelizable error-correcting or discrete-logarithm operations, we couple a strictly sequential unknown-order TLP core with the monolithic A-Scheme over a scrambled non-systematic Reed–Solomon code $\mathcal{C}_{\mathrm{RS}}[256, 192, 65]_{256}$. We formalize Time-Locked Noise Stripping: the sender superimposes a time-locked noise vector $\boldsymbol{\xi}_{\text{lock}}$ onto the A-Scheme's ephemeral half-radius blind noise $\boldsymbol{\xi}_i$ ($w_\xi = 23$), pushing the pre-release wire error weight strictly beyond both the unique Berlekamp–Massey decoding radius ($\aleph = 31$) and the Guruswami--Sudan Johnson bound ($J(255, 192) = 34$). We prove that parallelized GPU/ASIC adversaries cannot bypass the sequential delay via erasure-support guessing or symbol-cancellation grinding, while after $\mathcal{T}$ sequential steps, stripping $\boldsymbol{\xi}_{\text{lock}}$ unlocks blind decoding \textit{exclusively} for recipients inside the authorized fuzzy Hamming ball ($d_{\mathrm{H}}(\mathbf{r}_S, \mathbf{r}_R) \le 8$), leaving out-of-cohort solvers ($d_{\mathrm{H}} = 64$) isolated behind a $> 312$-bit statistical barrier. Finally, combining Elligator Squared (with explicit inverse isogenies) and constant-time inverse range sampling renders the entire FCTLP wire envelope indistinguishable from uniform noise ($\Delta(\mathcal{P}_\Pi, \mathrm{U}) \le 2^{-\lambda}$).
Authors
- Andrey Chmora
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-19
- DOI
- https://doi.org/10.5281/zenodo.24380778
- Primary Topic
- Cryptography and Data Security
- Type
- preprint