Provable Watermark Extraction

Tomer Solberg

Track: long paper (up to 9 pages)

Keywords: Watermark, CNN, Zero Knowledge, Privacy

TL;DR: Introducing zkDL++, a novel framework designed for provable AI.

Abstract:

Introducing zkDL++, a novel framework designed for provable AI. Leveraging zkDL++, we
address a key challenge in generative AI watermarking—maintaining privacy while
ensuring provability. By enhancing the watermarking system developed by Meta, zkDL++
solves the problem of needing to keep watermark extractors private to avoid attacks,
offering a more secure solution. Beyond watermarking, zkDL++ proves the integrity of any
deep neural network (DNN) with high efficiency. In this post, we outline our approach,
evaluate its performance, and propose avenues for further optimization.

Presenter: ~Tomer_Solberg1

Format: Yes, the presenting author will definitely attend in person because they are attending ICLR for other complementary reasons.

Funding: No, the presenting author of this submission does *not* fall under ICLR’s funding aims, or has sufficient alternate funding.

Anonymization: This submission has been anonymized for double-blind review via the removal of identifying information such as names, affiliations, and identifying URLs.

Submission Number: 15