20240005651. ADVERSARIAL DETECTION USING DISCRIMINATOR MODEL OF GENERATIVE ADVERSARIAL NETWORK ARCHITECTURE simplified abstract (INTUIT INC.)

From WikiPatents
Jump to navigation Jump to search

ADVERSARIAL DETECTION USING DISCRIMINATOR MODEL OF GENERATIVE ADVERSARIAL NETWORK ARCHITECTURE

Organization Name

INTUIT INC.

Inventor(s)

Miriam Hanna Manevitz of Hasharon (IL)

Aviv Ben Arie of Hasharon (IL)

ADVERSARIAL DETECTION USING DISCRIMINATOR MODEL OF GENERATIVE ADVERSARIAL NETWORK ARCHITECTURE - A simplified explanation of the abstract

This abstract first appeared for US patent application 20240005651 titled 'ADVERSARIAL DETECTION USING DISCRIMINATOR MODEL OF GENERATIVE ADVERSARIAL NETWORK ARCHITECTURE

Simplified Explanation

The patent application describes a method for training and deploying a generative adversarial network (GAN) in a production system. Here are the key points:

  • The method involves training a GAN using real data objects to create a generator model that can generate realistic data.
  • The GAN consists of a generator model and a discriminator model, which are trained together in an adversarial manner.
  • The discriminator model is trained using both adversarial data objects and additional real data objects to classify the authenticity of the data.
  • Once trained, the discriminator model is deployed to a production system.
  • In the production system, the discriminator model outputs an authenticity binary class for the data objects, which is then passed to a system classifier model.

Potential applications of this technology:

  • Data generation: The trained generator model can be used to generate realistic data for various applications, such as image synthesis or text generation.
  • Anomaly detection: The discriminator model can be used to classify data objects as authentic or adversarial, enabling the detection of anomalies or fraudulent activities.
  • Quality control: The authenticity binary class output by the discriminator model can be used to assess the quality of data objects in a production system.

Problems solved by this technology:

  • Generating realistic data: The GAN training process allows for the creation of a generator model that can produce data objects that closely resemble real data.
  • Adversarial detection: The discriminator model is trained to distinguish between authentic and adversarial data objects, providing a means to identify potential threats or anomalies.
  • Scalability: The trained discriminator model can be deployed in a production system to process large volumes of data and provide real-time authenticity classification.

Benefits of this technology:

  • Improved data generation: The trained generator model can generate high-quality data that can be used for various purposes, such as training machine learning models or conducting simulations.
  • Enhanced security: The discriminator model can help identify and mitigate potential adversarial attacks or fraudulent activities in a production system.
  • Real-time classification: The deployed discriminator model can provide fast and accurate authenticity classification for data objects, enabling timely decision-making and response.


Original Abstract Submitted

a method includes training, using first real data objects, a generative adversarial network having a generator model and a discriminator model to create a trained generator model that generates realistic data, and training, using adversarial data objects and second real data objects, the discriminator model to output an authenticity binary class for the adversarial data objects and the second real data objects. the method further includes deploying the discriminator model to a production system. in the production system, the discriminator model outputs the authenticity binary class to a system classifier model.

Retrieved from "http://wikipatents.org/index.php?title=20240005651._ADVERSARIAL_DETECTION_USING_DISCRIMINATOR_MODEL_OF_GENERATIVE_ADVERSARIAL_NETWORK_ARCHITECTURE_simplified_abstract_(INTUIT_INC.)&oldid=26998"