Adversarial Neuro Encoding with Binary Neural Networks

Adversarial neuro encoding could provide new insights for ciphering information with different perspectives. Nevertheless, it is still underexplored with a handful of publications on the subject. This work proposes the implementation of neuroevolved binary neural networks based on boolean logic func...

Full description

Saved in:
Bibliographic Details
Published in2020 IEEE Asia-Pacific Conference on Computer Science and Data Engineering (CSDE) pp. 1 - 5
Main Authors Valencia, Raul, Sham, Chiu Wing
Format Conference Proceeding
LanguageEnglish
Published IEEE 16.12.2020
Subjects
Adversarial Neurocryptography
Binary Neural Network
BiSUNA
CPA
Encoding
Image coding
Neural networks
Neuroevolution
Reinforcement learning
Sociology
Training
Visualization
Online AccessGet full text
DOI10.1109/CSDE50874.2020.9411537

Cover

More Information
Summary:Adversarial neuro encoding could provide new insights for ciphering information with different perspectives. Nevertheless, it is still underexplored with a handful of publications on the subject. This work proposes the implementation of neuroevolved binary neural networks based on boolean logic functions only (BiSUNA) that apply payload ciphering between two agents to disperse information from an observer. The BiSUNA framework provides three distinctive attributions: it uses an adversarial neural encoding environment to improve the system data transmission; one execution yields a diversity of results given its population heuristics; lastly, it is an unconventional proposal to employ binary neural networks for the solution of symmetric ciphered problems.
DOI:10.1109/CSDE50874.2020.9411537