Work

Preprints, articles and reports. When one version corrects another I say so, and when a result is a conjecture or an extrapolation, I say that too.

Type

Year

  1. AURA: An Uncertainty-Aware Architecture for Verifiable Decisions in Critical Infrastructure

    Preprint, · v3 · Lucas Alaniz Pintos

    Supersedes v1 (October 2025), which described deployment results, validation studies and external evaluations that had not been carried out. v3 reports only what is implemented and reproducible.

    A three-layer architecture, which I developed at Ineco, for operational decisions under contradictory weather forecasts. It fuses heterogeneous sources into credal sets with the theory of Dempster and Shafer, propagates the uncertainty through a structural causal model with declared per-equation error bounds, and certifies decisions against safety specifications. Under four explicit, falsifiable hypotheses (H0 to H3), the computed sets contain the true state end to end.

    • The verification artifact includes the code and 546 tests.
  2. Quantum walks on configuration matroids revisited, with a corrected speedup and a rigorous classical comparison

    Preprint, · v2 · Lucas Alaniz Pintos

    Corrects and supersedes v1, from December 2025. It withdraws the claim of a speedup growing as √n with network size and includes an appendix with 14 errata.

    Restates the framework on the down-up chain of the cographic dual, with a corrected cost T_Q = O(√(k|T|/|M|)) against T_C = O(k|T|/|M|). Against classical competitors on 22 operational marked sets, the quantum method is cheaper in AC power-flow solves in 0 of 44 comparisons. What survives is a query-complexity separation against the chain’s own random walk, with an executable gate-level oracle.

  3. Exact spectral gap of the down-up walk on matroids with parallel elements

    Preprint, · Lucas Alaniz Pintos

    For a matroid of rank r, the down-up walk has spectral gap in [1/r, 2/r]. The note proves that any pair of parallel elements pins the gap at exactly 1/r, which for cographic matroids covers every graph with a degree-two vertex, and that the uniform matroids U(r, 2r−2) attain the 2/r ceiling. Both ends of the interval are realised.

  4. Placement is a Primitive

    Preprint, · Lucas Alaniz Pintos

    Identifies a positional design lever for arithmetization-oriented SPN hash functions over the Goldilocks field. Folding a quadratic coupling into the S-box input adds one bit of CICO ideal degree per round, while placing it in the linear layer or after the S-box adds nothing. msolve measurements indicate the effect is genuine and generic. The principle is instantiated as Alaniz-AO, a Goldilocks sponge.

    • Round counts and cost figures are extrapolations from the measured degree law under an explicit cost model (ω = 2). They are not measurements.
  5. Calibrated-Uncertainty Detection for Counter-UAS: Order-Invariant Evidential Fusion with Guaranteed-Containment Tracking

    Preprint, · Lucas Alaniz Pintos

    The preprint of AEGIS, a detection-only counter-drone system I developed at Ineco. It tells drones apart from birds, clutter and empty scenes under noisy and sometimes contradictory evidence, and under genuine conflict it abstains instead of giving a confident wrong answer.

    • I developed it at Ineco. Detection only, with a human in the loop and no countermeasure capability.
  6. The Alaniz Cipher v3: Sheaf-Based Post-Quantum Encryption via Vector-Valued Finite Field Permutations

    Preprint, · v3 · Lucas Alaniz Pintos

    Fixes v1 (broken by A. Rodríguez Langa with a scaling attack) and shows the break of v2.

    A public-key encryption scheme built from cellular sheaves on graphs with nonlinear restriction morphisms over extension fields. Encryption becomes probabilistic with a nonce-derived stream (SHAKE-256) and σ acts vectorially in F_{p^d}. The paper also presents a chosen-plaintext key-recovery attack on v2.

    • It gives no formal worst-case to average-case reduction for NL-SMIP, a limitation shared by multivariate schemes.
    • The paper’s PQ-128 security estimate was later revisited in the v4 repository, and it does not hold.
  7. Projection-Valued Lorentzian Polynomials and an Operator Hodge–Riemann Theory for Matroids

    Preprint, · Lucas Alaniz Pintos

    Introduces projection-valued Lorentzian polynomials, an operator-valued extension of the Lorentzian polynomials of Brändén and Huh. For loopless rank-2 matroids it states an operator Hodge-Riemann theorem with a universal signature formula for the operator Hessian, and adds a diagonal normalisation identity, a Szegedy-type spectral correspondence and contraction reduction to rank 2.

    • The signature formula is proved only in the regular case (all c_i equal). The general result is a conjecture.
  8. AURA Port – Sistema de Verificación Formal para Operaciones Portuarias

    Report, · Lucas Alaniz Pintos

    Extends the AURA framework to critical port operations. It combines multi-source Bayesian fusion, contextual anomaly detection and formal verification with PVS, and integrates weather and ocean data (AEMET, NASA POWER, Copernicus, REDEXT buoys) to compute risk indices and issue auditable safety certificates. The report is in Spanish.

    • Derived from AURA v1. The operational-validation claims in this report are superseded by AURA v3, which reports only what is implemented and reproducible.
  9. AURA Wildfire – Autonomous Resilience System for Forest Fire Scenarios

    Report, · Lucas Alaniz Pintos

    Extends the AURA framework to forest fires, with a three-layer architecture of probabilistic fusion, causal modelling and formal decision verification under extreme weather uncertainty. It integrates remote sensing (NASA FIRMS, Copernicus Sentinel-2/3, ERA5, AEMET), physical fire models (FWI, Rothermel, Byram) and automated theorem proving with PVS. The report is in Spanish.

    • A conceptual evaluation by simulation with synthetic data. It shows theoretical viability, not operational results.
  10. Modelado de algoritmos cuánticos para la simulación de redes neuronales biológicas

    Article, · Lucas Alaniz Pintos

    Metrópolis | Revista de Estudios Universitarios Globales, vol. 5, no. 1, pp. 102-114

    Explores the potential of quantum algorithms for simulating biological neural networks, using IBM Quantum cloud computing and artificial-intelligence techniques to optimise circuit parameters. It works with connectivity and gene-expression data from the Allen Brain Atlas, processed for use in quantum simulations, and details the steps, results and conclusions of the process. The article is in Spanish.