arXiv subjects

David Chernin

Publications and source records attributed to David Chernin.

4 recordsLinked to original sources

arXiv2026

CRAW: Codec Robust Audio Watermarking

Recent advances in generative speech models have made it increasingly difficult to distinguish authentic from synthetic audio, enabling new forms of fraud and misinformation. Audio watermarking offers a promising defense by embedding an imperceptible signal into generated speech that can later be detected to verify its provenance. However, recent studies have shown that existing post-hoc watermarking methods fail under neural codecs and denoisers, transformations routinely applied during real-world storage, transmission, and processing, severely limiting their practical utility. Here we introduce CRAW, a codec-robust audio watermarking framework that jointly improves robustness against neural re-synthesis while maintaining high perceptual quality. CRAW combines distortion-aware training with an attention-based pooling mechanism, inference-time perceptual mask- ing, and an error-correcting code to recover the fidelity lost during robust training. Experiments demonstrate that CRAW achieves state-of-the-art robustness against neural codecs, denoisers, and vocoders while maintaining perceptual quality comparable to existing post-hoc watermarking methods. The code is available at https://github.com/DavidC1212/craw.

cs.SD↗

arXiv2025

Few-Shot Speech Deepfake Detection Adaptation with Gaussian Processes

Recent advancements in Text-to-Speech (TTS) models, particularly in voice cloning, have intensified the demand for adaptable and efficient deepfake detection methods. As TTS systems continue to evolve, detection models must be able to efficiently adapt to previously unseen generation models with minimal data. This paper introduces ADD-GP, a few-shot adaptive framework based on a Gaussian Process (GP) classifier for Audio Deepfake Detection (ADD). We show how the combination of a powerful deep embedding model with the Gaussian processes flexibility can achieve strong performance and adaptability. Additionally, we show this approach can also be used for personalized detection, with greater robustness to new TTS models and one-shot adaptability. To support our evaluation, a benchmark dataset is constructed for this task using new state-of-the-art voice cloning models.

cs.SD↗

arXiv2023

Limiting Current on Periodic Electron Sheets in a Planar Diode

We consider the steady state limiting current that can be carried by an infinite periodic array of thin electron sheets spaced by period p in a planar diode of gap voltage V and gap separation d. Our primary assumptions are (1) electron motion is restricted by an infinite magnetic field to the direction normal to the electrode surfaces, (2) all electrons are emitted from the cathode with initial kinetic energy Ein, and (3) electron motion is non-relativistic. The limiting current density, averaged over a period and normalized to the classical 1D Child-Langmuir (CL) current density (including a factor that accounts for non-zero Ein), is found to depend only on the two dimensionless parameters p/d and Ein/eV. This average limiting current density is computed from the maximum current density for which the iterative solution of a non-linear integral equation converges. Numerical results and empirical curve fits for the limiting current are presented, together with an analysis as p/d and Ein/eV approach zero or infinity, in which cases previously published results are recovered. Our main finding is that, while the local anode current density within each electron sheet is infinite in our model (that is, it exceeds the classical 1D CL value by an 'infinite' factor), the period average anode current density is in fact still bounded by the classical 1D CL value. This study therefore provides further evidence that the classical 1D Child-Langmuir current density is truly a fundamental limit that cannot be circumvented.

physics.plasm-ph↗

arXiv2018

Adjoint Approach to Beam Optics Sensitivity Based on Hamiltonian Particle Dynamics

We develop a sensitivity function for the design of electron optics using an adjoint approach based on a form of reciprocity implicit in Hamilton's equations of motion. The sensitivity function, which is computed with a small number of time-reversed runs of a beam optics code, allows for the determination of the effects on specific beam quality figures of merit of small, but arbitrary changes in electrode potentials, positions and shapes, and in magnet strengths and locations. The sensitivity function can thus be used in an optimization cycle of a focusing system's design, and/or to predict the sensitivity of a particular design to manufacturing, assembly, and alignment errors.

physics.acc-ph↗