Chaos-based intelligent model black-box watermark trigger set automatic labeling algorithm

An automatic labeling and chaotic technology, applied in the field of image processing, can solve problems such as inability to find trigger sets, non-generalization, and inability to predict chaotic characteristics, to ensure non-generalizability, commercial savings, and time and labor savings. Effect

Active Publication Date: 2022-06-03
XIAMEN UNIV TAN KAH KEE COLLEGE
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Machine learning cannot predict the characteristics of chaos, and the attacker cannot find other trigger sets that match the characteristics of our watermark
Therefore, the watermark is not generalizable

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Chaos-based intelligent model black-box watermark trigger set automatic labeling algorithm
  • Chaos-based intelligent model black-box watermark trigger set automatic labeling algorithm
  • Chaos-based intelligent model black-box watermark trigger set automatic labeling algorithm

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0029] Below in conjunction with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely

[0038] The 32 images marked by the chaotic automatic labeler can be regarded as a trigger set, and the training set is collected before.

[0039] The experimental and simulation results show that our automatic labeling algorithm is resistant to fine-tuning, compression, overlay and other attacks.

[0041] Fine-tuning and transfer learning are the most common attacks because of the heavy workload of training a model from scratch. I

[0043] Compression attack is also a common attack method. We use the TensorFlow model optimization tool to prune i

[0045]

[0047] The overlay attack assumes that the attacker knows our watermarking mechanism. Then, the adversary according to our watermark generation algorithm

[0048] By changing the parameters and initial values ​​of the logisti...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses an intelligent model black-box watermark trigger set automatic labeling algorithm based on chaos. This technology uses chaos to solve the problem of automatic labeling of trigger set tags, and can effectively solve the problem of automatic labeling of trigger sets for existing black-box watermarks, so that we can use remote APIs to extract watermarks with a small number of queries, and then determine all of the deep learning intelligent models. identity. It can effectively solve the problem of difficult extraction of watermark by white-box watermarking scheme, without downloading the server-side model or reading the source program. Judicial forensics will also become extremely simple. It is only necessary to compare whether the output classification results are consistent with the expected results.

Description

Chaos-based intelligent model black-box watermark trigger set automatic labeling algorithm technical field The present invention relates to the technical field of image processing, be specifically based on the intelligent model black box watermark trigger set automatic of chaos Labeling algorithm. Background technique [0002] At present, in the development of a new generation of artificial intelligence in my country, the Ministry of Science and Technology has initiated the construction of a new generation of artificial intelligence in China. Put the innovation platform, and the typical ones established are: "autonomous driving (Baidu), intelligent vision (shangtang group), medical imaging (Tengzhou) News), intelligent voice (iFLYTEK), etc.". The creation of such high-performance deep learning intelligent models requires a lot of money manpower and material resources, and they have been deeply integrated with human society. Deep learning intelligence models can be vie...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Patents(China)
IPC IPC(8): G06T1/00
CPCG06T1/005G06T2201/0065
Inventor 张盈谦贾贻然闻芊芊
Owner XIAMEN UNIV TAN KAH KEE COLLEGE
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap