Vishal Asnani

Vishal Asnani

Research Scientist at Adobe

Trustworthy AI · Watermarking · Media provenance

Posts by Collection

portfolio

publications

Dual Band Microstrip Patch Antenna with Secondary Patch

Published in IEEE International Conference on Communication and Signal Processing (ICCSP), 2018

Can a compact printed antenna support two separated operating bands? This earlier research studies how patch rotation and an additional feed-connected patch change impedance matching.

Recommended citation: Sudeep Baudha, Vishal Asnani. (2018). Dual Band Microstrip Patch Antenna with Secondary Patch. https://doi.org/10.1109/ICCSP.2018.8524225

Proactive Image Manipulation Detection

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022

Can images be prepared to make future manipulation easier to detect? This work moves part of the forensic process before the edit: a learned template is added to an original image, giving the detector a signal to examine later.

Recommended citation: Asnani, V., Yin, X., Hassner, T., Liu, S., & Liu, X. (2022). Proactive Image Manipulation Detection. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. https://openaccess.thecvf.com/content/CVPR2022/html/Asnani_Proactive_Image_Manipulation_Detection_CVPR_2022_paper.html

MaLP: Manipulation Localization Using a Proactive Scheme

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023

Knowing an image is fake does not tell us where it changed. MaLP adds a learned template before manipulation and uses that signal to estimate a pixel-level fakeness map.

Recommended citation: Asnani, V., Yin, X., Hassner, T., & Liu, X. (2023). MaLP: Manipulation Localization Using a Proactive Scheme. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. https://openaccess.thecvf.com/content/CVPR2023/html/Asnani_MaLP_Manipulation_Localization_Using_a_Proactive_Scheme_CVPR_2023_paper.html

PrObeD: Proactive Object Detection Wrapper

Published in Thirty-seventh Conference on Neural Information Processing Systems 2023, 2023

Instead of only changing the detector, can we also learn a better input? PrObeD wraps object detectors with an image-dependent template generator, aiming to make useful scene information easier to exploit.

Recommended citation: Asnani V., Kumar A., You S., Liu X., (2023). PrObeD: Proactive Object Detection Wrapper. In Thirty-seventh Conference on Neural Information Processing Systems. https://proceedings.neurips.cc/paper_files/paper/2023/hash/f5846131aa6a72d1df3bd6d43a4a960b-Abstract-Conference.html

Reverse Engineering of Generative Models: Inferring Model Hyperparameters from Generated Images

Published in IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

What can a generated image reveal about the model behind it? This work defines model parsing: estimating generator architecture and training-loss attributes from generated images, including images from models unseen during training.

Recommended citation: Asnani, V., Yin, X., Hassner, T., & Liu, X. (2023). Reverse engineering of generative models: Inferring model hyperparameters from generated images. Transactions on Pattern Analysis and Machine Intelligence, vol. 45, no. 12, pp. 15477-15493. https://doi.org/10.1109/TPAMI.2023.3301451

Reverse Engineering of Deceptions on Machine- and Human-Centric Attacks

Published in Foundations and Trends in Privacy and Security, 2024

Detecting deception is only one part of understanding it. This tutorial monograph connects attacks on machine-learning systems with synthetic media intended to mislead people, asking what their outputs reveal about the underlying toolchain.

Recommended citation: Yao, Y., Guo, X., Asnani, V., Gong, Y., Liu, J., Lin, X., ... & Liu, S. (2024). Reverse Engineering of Deceptions on Machine-and Human-Centric Attacks. Foundations and Trends® in Privacy and Security, 6(2), 53-152. https://doi.org/10.1561/3300000039

ProMark: Proactive Diffusion Watermarking for Causal Attribution

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024

Visual resemblance alone does not establish which training concepts influenced a generated image. ProMark introduces concept-specific watermark signals during training and retrieves those signals from generated content.

Recommended citation: Asnani, V., Collomosse, J., Bui, T., Liu, X., Agarwal, S. (2024). ProMark: Proactive Diffusion Watermarking for Causal Attribution. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. https://openaccess.thecvf.com/content/CVPR2024/html/Asnani_ProMark_Proactive_Diffusion_Watermarking_for_Causal_Attribution_CVPR_2024_paper.html

Tracing Hyperparameter Dependencies for Model Parsing via Learnable Graph Pooling Network

Published in Advances in Neural Information Processing Systems (NeurIPS), 2024

Generator attributes are not independent: architecture and training choices often have structure. This work uses that structure to improve model parsing from generated images.

Recommended citation: Xiao Guo, Vishal Asnani, Sijia Liu, Xiaoming Liu. (2024). Tracing Hyperparameter Dependencies for Model Parsing via Learnable Graph Pooling Network. https://proceedings.neurips.cc/paper_files/paper/2024/hash/d412468559da4e2ac41a0285b3a079cd-Abstract-Conference.html

CustomMark: Customization of Diffusion Models for Proactive Attribution

Published in IEEE/CVF International Conference on Computer Vision, APAI Workshop (ICCV Workshop), 2025

Can attribution be added to a pretrained generator without retraining the full model? CustomMark selectively customizes concepts so that generated images carry identifiers for relevant objects, styles, or other concepts.

Recommended citation: Vishal Asnani, John Collomosse, Xiaoming Liu, Shruti Agarwal. (2025). CustomMark: Customization of Diffusion Models for Proactive Attribution. https://openaccess.thecvf.com/content/ICCV2025W/APAI/html/Asnani_CustomMark_Customization_of_Diffusion_Models_for_Proactive_Attribution_ICCVW_2025_paper.html

Your Text Encoder Can Be An Object-Level Watermarking Controller

Published in IEEE/CVF International Conference on Computer Vision (ICCV), 2025

Watermarking need not apply uniformly to a whole generated image. This work learns a text token that can control a watermark globally or steer it toward selected objects.

Recommended citation: Naresh Kumar Devulapally, Mingzhen Huang, Vishal Asnani, Shruti Agarwal, Siwei Lyu, Vishnu Suresh Lokhande. (2025). Your Text Encoder Can Be An Object-Level Watermarking Controller. https://openaccess.thecvf.com/content/ICCV2025/html/Devulapally_Your_Text_Encoder_Can_Be_An_Object-Level_Watermarking_Controller_ICCV_2025_paper.html

NRGMark: Localized Watermarking for Energy Transparency in Images

Published in IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2026

A composite design can contain several images with different origins and energy footprints. NRGMark gives individual visual elements independent identifiers rather than assigning one provenance payload to the whole design.

Recommended citation: Shruti Agarwal, Elie Michel, Vishal Asnani, Tania Mathern, John Collomosse. (2026). NRGMark: Localized Watermarking for Energy Transparency in Images. https://openaccess.thecvf.com/content/WACV2026/html/Agarwal_NRGMark_Localized_Watermarking_for_Energy_Transparency_in_Images_WACV_2026_paper.html

Proactive Schemes: A Survey of Adversarial Attacks for Social Good

Published in International Journal of Computer Vision (IJCV), 134(4), 2026

The same tools used to construct adversarial perturbations can also support protective and constructive uses. This survey organizes proactive schemes that deliberately modify inputs to help a downstream task.

Recommended citation: Asnani, V., Yin, X., Liu, X. (2026). Proactive Schemes: A Survey of Adversarial Attacks for Social Good. International Journal of Computer Vision, 134(4). https://doi.org/10.1007/s11263-025-02631-z

What Do Deepfake Benchmarks Measure? An Audit Using Frozen Self-Supervised Representations

Preprint, 2026

Does a high deepfake-detection score necessarily demonstrate forensic understanding? This audit asks how much benchmark performance is already available in general-purpose, frozen representations.

Recommended citation: Samuel Pagon, Yixuan Shen, Vishal Asnani, Feng Liu. (2026). What Do Deepfake Benchmarks Measure? An Audit Using Frozen Self-Supervised Representations. arXiv:2606.26384. https://arxiv.org/abs/2606.26384

Interpretable Prompts made Edit-Friendly: Token-to-Token Similarity Reduction in dLLMs for Edit-Friendly Hard Prompt Inversion

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

An image-reconstruction prompt is more useful if a person can understand and edit it. This work inverts reference images into discrete prompts designed for later word substitutions and additions.

Recommended citation: Naresh Kumar Devulapally, Shruti Agarwal, Vishal Asnani, Vishnu Suresh Lokhande. (2026). Interpretable Prompts made Edit-Friendly: Token-to-Token Similarity Reduction in dLLMs for Edit-Friendly Hard Prompt Inversion. https://openaccess.thecvf.com/content/CVPR2026/html/Devulapally_Interpretable_Prompts_made_Edit-Friendly_Token-to-Token_Similarity_Reduction_in_dLLMs_for_CVPR_2026_paper.html

PiVoT: Proactive Video Templates for Enhancing Video Task Performance

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2026

Can a learned change to video frames make motion easier for a model to understand? PiVoT extends proactive templates to video, pairing temporal input adaptation with parameter-efficient model adaptation.

Recommended citation: Vishal Asnani, Shruti Agarwal, Xiaoming Liu. (2026). PiVoT: Proactive Video Templates for Enhancing Video Task Performance. https://openaccess.thecvf.com/content/CVPR2026W/CV4Smalls/html/Asnani_PiVoT_Proactive_Video_Templates_for_Enhancing_Video_Task_Performance_CVPRW_2026_paper.html

TokenTrace: Multi-Concept Attribution through Watermarked Token Recovery

Published in IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2026

One generated image can contain several creative influences. TokenTrace asks how to verify each influence separately, rather than treating the entire image as one attribution target.

Recommended citation: Li Zhang, Shruti Agarwal, John Collomosse, Pengtao Xie, Vishal Asnani. (2026). TokenTrace: Multi-Concept Attribution through Watermarked Token Recovery. https://openaccess.thecvf.com/content/CVPR2026/html/Zhang_TokenTrace_Multi-Concept_Attribution_through_Watermarked_Token_Recovery_CVPR_2026_paper.html

FlowMark: Mask-Guided Video Watermarking

Preprint, 2026

Where should a watermark go in a video frame? FlowMark learns content-adaptive embedding regions rather than relying on manually supplied or random masks. The goal is to preserve visual quality and temporal consistency while keeping provenance signals recoverable.

Recommended citation: Vishal Asnani, Shruti Agarwal, John Collomosse. (2026). FlowMark: Mask-Guided Video Watermarking. arXiv:2607.05261. https://arxiv.org/abs/2607.05261

Signpost Watermarking: Joint Optimization for Visual Watermark Coexistence

Accepted to ECCV 2026 LifegenIP Workshop, 2026

How can one watermark help discover another without disrupting it? Signpost watermarking investigates a lightweight routing signal that indicates which independently deployed provenance watermark should be decoded.

Recommended citation: Shruti Agarwal, Vishal Asnani, John Collomosse. (2026). Signpost Watermarking: Joint Optimization for Visual Watermark Coexistence. Accepted to the ECCV 2026 LifegenIP Workshop. https://arxiv.org/abs/2608.10091

talks

Reverse Engineering of Deceptions

Published:

Co-organized a tutorial on reverse engineering of deceptions, adversarial machine learning, and trustworthy computer vision.

Foundations and Frontiers of Watermarking

Published:

Co-organized a half-day tutorial on classical and deep-learning watermarking, multimodal and generative watermarking, robustness benchmarks, and content authenticity frameworks including C2PA and Content Credentials.

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.