AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

Hashed Watermark as a Filter: A Unified Defense Against Forging and Overwriting Attacks in Neural Network Watermarking

Yuan Yao, Jin Song, Jian Jin

AAAI 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Neural network watermarking

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use hashed watermark as a parameter filterNeuralMark · CIFAR-10/100 forging attack, ResNet-18counterfeit detection rate (%): 100.00 (VanillaMark) -> 48.56Improved
use hashed watermark as a filter under strong overwriting (lambda 1000)NeuralMark · CIFAR-100 to CIFAR-10, ResNet-18original watermark detection rate (%): 53.90 (VanillaMark) -> 100Improved
use hashed watermark filter at lr 0.005 overwritingNeuralMark · CIFAR-100 to CIFAR-10, ResNet-18original watermark detection rate (%): 73.04 (VanillaMark) -> 99.60Improved
increase overwriting learning rate to 0.1 or moreNeuralMark · CIFAR-100 to CIFAR-10, ResNet-18detection rate and accuracy: detection 50.78, accuracy 75.2 at lr 0.1Hurt
add average pooling to filtered parametersNeuralMark · fine-tuning and pruning attackswatermark detection rate: 100% under fine-tuningImproved
embed watermark with NeuralMarkNeuralMark · CIFAR-100, 9 architecturesaccuracy (%): WideResNet-50 59.67 -> 58.41; ViT-B/16 39.07 -> 39.22No clear effect
embed watermark with NeuralMarkNeuralMark · E2E text generationBLEU: GPT-2-S 69.36 -> 69.59; GPT-2-M 68.7 -> 67.73Mixed
increase number of filtering rounds (ablation)hashed watermark filter · CIFAR-100 ResNet-18parameter overlap ratioImproved
use 6 or 8 filtering rounds instead of 4 (ablation)NeuralMark · robustness against attacksdetection rateNo clear effect
use a plain private filter instead of hashed watermark filter (ablation)private-filter baseline · forging attackforging successHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy