Sept. 20, 2023, 1:10 a.m. | Weina Dong, Jia Liu, Yan Ke, Lifeng Chen, Wenquan Sun, Xiaozhong Pan

cs.CR updates on arXiv.org arxiv.org

The utilization of implicit representation for visual data (such as images,
videos, and 3D models) has recently gained significant attention in computer
vision research. In this letter, we propose a novel model steganography scheme
with implicit neural representation. The message sender leverages Neural
Radiance Fields (NeRF) and its viewpoint synthesis capabilities by introducing
a viewpoint as a key. The NeRF model generates a secret viewpoint image, which
serves as a backdoor. Subsequently, we train a message extractor using
overfitting to …

attention backdooring capabilities computer computer vision data images letter message novel representation research steganography videos

CyberSOC Technical Lead

@ Integrity360 | Sandyford, Dublin, Ireland

Cyber Security Strategy Consultant

@ Capco | New York City

Cyber Security Senior Consultant

@ Capco | Chicago, IL

Sr. Product Manager

@ MixMode | Remote, US

Security Compliance Strategist

@ Grab | Petaling Jaya, Malaysia

Cloud Security Architect, Lead

@ Booz Allen Hamilton | USA, VA, McLean (1500 Tysons McLean Dr)