Robustness Studies in 3D Camera Data

Abstract

Our main aim is the iterative development of Robust 3D (R3D) framework creating more sophisticated attacks to tamper the 3D camera data and improving the forensic techniques to detect such attacks. Towards this end, we will: (a)Design an anti-forensic framework that will carry out attacks on the 3D camera data stream by using different techniques as appropriate for the nature of the generated 3D data; (b) Next, use existing digital forensic techniques to analyze the manipulated 3D camera data stream and authenticate the data validity. The R3D framework comprises Anti-forensic 3D object stream manipulation framework: to capture and manipulate or attack 3D data streams to generate fake streams in such a way that it is difficult to detect the manipulations. This anti-forensic framework uses different techniques depending on the type of 3D camera. For data generated by LiDAR cameras, R3D introduces attacks such as additive (copy-paste, etc) and subtractive forgeries. For RGB-D streams, R3D can use a forgery skeleton sequence from a different source as inputs. The forged skeleton sequence can come from another live/recorded stream or one created using animation software, such as the Autodesk Motion builder. The system then produces a real-time realistic sequence of 3D models, like a 3D reconstruction system, would, but with the actor in the live stream performing actions shown by the forged skeleton sequence. This anti-forensic framework can also introduce human(s) into a scene where there were no humans. Forensic Framework: can analyze the 3D camera data streams and determine the possibilities for forgeries. This framework also uses techniques that vary with the type of camera. For LiDAR cameras, it uses techniques such as density consistency analyses. For RGB-D data streams, it uses approaches such as block-based depth noise evaluation to detect manipulations in the depth stream.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jan 30, 2023
Accession Number
AD1210544

Entities

People

  • Gopal Gupta

Organizations

  • University of Texas at Dallas

Tags

Communities of Interest

  • Autonomy
  • Cyber

DTIC Thesaurus Topics

  • Artificial Intelligence
  • Artificial Intelligence Software
  • Augmented Reality
  • Autonomous Systems
  • Computational Science
  • Computers
  • Health Services
  • Human Systems Integration
  • Information Systems
  • Jointsanatomy
  • Machine Learning
  • Medical Personnel
  • Mixed Reality
  • Motion Capture
  • Neural Networks
  • Rehabilitation
  • Robot Navigation
  • Robots
  • Virtual Reality

Fields of Study

  • Computer science

Readers

  • Computer Vision.
  • Cybersecurity.