arXiv subjects

Thuc D. Nguyen

Publications and source records attributed to Thuc D. Nguyen.

3 recordsLinked to original sources

arXiv2026

Transferable Multi-Bit Watermarking Across Frozen Diffusion Models via Latent Consistency Bridges

As generative AI advances, global governance frameworks increasingly mandate verifiable content provenance. However, existing watermarking techniques face a critical policy-to-technology disconnect: sampling-based methods require computationally prohibitive inversion, while fine-tuning approaches are tethered to specific model checkpoints, hindering standardized, cross-model oversight. To bridge this gap, we introduce DiffMark, a plug-and-play multi-bit watermarking framework. DiffMark embeds a persistent, learned perturbation into every denoising step of a frozen diffusion model, accumulating a recoverable signal in the final latent space. To enable efficient training through the frozen network, we utilize Latent Consistency Models (LCMs) as a differentiable training bridge. DiffMark achieves 64-bit extraction in a single 16.4 ms forward pass, which is a $45\times$ speed-up over inversion baselines. By enabling per-image key flexibility and cross-architecture transferability without retraining, DiffMark provides the practical, scalable technical tooling necessary to operationalize user accountability and enforce emerging AI governance mandates.

cs.CV↗

arXiv2023

New results on Erasure Combinatorial Batch Codes

We investigate in this work the problem of Erasure Combinatorial Batch Codes, in which $n$ files are stored on $m$ servers so that every set of $n-r$ servers allows a client to retrieve at most $k$ distinct files by downloading at most $t$ files from each server. Previous studies have solved this problem for the special case of $t=1$ using Combinatorial Batch Codes. We tackle the general case $t \geq 1$ using a generalization of Hall's theorem. Additionally, we address a realistic scenario in which the retrieved files are consecutive according to some order and provide a simple and optimal solution for this case.

cs.IT↗

arXiv2021

Improved algorithms for non-adaptive group testing with consecutive positives

The goal of group testing is to efficiently identify a few specific items, called positives, in a large population of items via tests. A test is an action on a subset of items which returns positive if the subset contains at least one positive and negative otherwise. In non-adaptive group testing, all tests are fixed in advance and can be performed in parallel. In this work, we consider non-adaptive group testing with consecutive positives in which the items are linearly ordered and the positives are consecutive in that order. We present two contributions here. The first is the direct use of a binary code to construct measurement matrices compared to the use of Gray code in the state-of-the-art work, which is a rearrangement of the binary code, when the maximum number of consecutive positives is known. This leads to a reduction in decoding time in practice. The second one is efficient designs to identify positives when the number of consecutive positives is known. To the best of our knowledge, this setting has not been surveyed yet. Our simulations verify the efficiency of our proposed designs. In particular, it only requires up to $300$ tests to identify up to $100$ positives in a set of $2^{32} \approx 4.3\mathrm{B}$ items in less than $300$ nanoseconds. When the maximum number of consecutive positives is known, the simulations validate the superiority of our proposed design in decoding compared to the state-of-the-art work. Moreover, when the number of consecutive positives is known, the number of tests and the decoding time are almost reduced half.

cs.IT↗