This application relates to a face image processing method, apparatus, computer device, and storage medium. The method includes acquiring a first face image and a second face image, the first face image and the second face image being images of real faces; generating a first updated face image with non-real face image characteristics based on the first face image; adjusting color distribution of the first updated face image according to color distribution of the second face image to obtain a first adjusted face image; acquiring a target face mask of the first face image, the target face mask being generated by randomly deforming a face region of the first face image; and blending the first adjusted face image and the second face image according to the target face mask to obtain a target face image. Accordingly, a diversity of target face images can be generated.

An information processing device 1 includes an assessment unit 12-2 that acquires information on an object of a user through a network N, and acquires a purchase price based on the information on the object at the time of purchasing the object, a contract unit 12-3 that gives a notification of a purchase price, a lease charge when leasing the object to be purchased to a user, and various conditions relating to a lease contract through the network, and accepts the lease contract, and a contract management unit 12-5 that pays the purchase price to a user, accepts payment of the lease charge, and manages a lease charge payment situation. A user can raise a fund corresponding to the purchase price with easy means of transmitting information on the object through the network N.

The present disclosure provides methods and systems for vehicle verification. The method may include receiving vehicle verification information related to a vehicle to be verified from a client, wherein the vehicle verification information includes a plurality of images acquired by the client via an imaging device, and the vehicle verification information responds to at least one vehicle verification instruction. The method may further include determining a verification result of the vehicle based on the vehicle verification information.

Systems and techniques for facilitating detection of data duplication issues relating to generation of non-fungible tokens are provided. In various embodiments, a computer system can access a digital artwork image. In various aspects, the computer system can generate a set of plagiarism probabilities by comparing the digital artwork image to a set of cached digital artwork images. In various instances, a given plagiarism probability in the set of plagiarism probabilities can indicate a likelihood that the digital artwork image was derived from a given cached digital artwork image in the set of cached digital artwork images. In various cases, the computer system can calculate an authenticity score for the digital artwork image based on the set of plagiarism probabilities. In various aspects, the computer system can determine whether the authenticity score for the digital artwork image satisfies a threshold authenticity value.

There is disclosed a method of creating a template of original video content, which is performed on a computer device that has access to a previously generated database of original video content. The method comprises receiving identifiers for at least a portion of an original video content; extracting at least a portion of metadata of the original video content; extracting at least a portion of frames from a sequence of frames of the original video content; identifying a sequence of scenes; creating a vector of the sequence of scenes; generating a template of the original video content that includes at least the portion of the metadata, and a vector of the sequence of scenes of the original video content; and storing the template in a database.

Embodiments of the present invention concern detecting Test-Time Evasion (TTE) attacks on neural network, particularly deep neural network (DNN), classifiers. The manner of detection is similar to that used to detect backdoors of a classifier whose training dataset was poisoned. Given knowledge of the classifier itself, the adversary subtly (even imperceptibly) perturbs their input to the classifier at test time in order to cause the class decision to change from a source class to a target class. For example, an image of a person who is unauthorized to access a resource can be modified slightly so that the classifier decides the image is that of an authorized person. The detector is based on employing a method (similar to that used to detect backdoors in DNNs) to discover different such minimal perturbations for each in a set of clean (correctly classified) samples, to change the sample's ground-truth (source) class to every other (target) class. For each (source, target) class pair, null distributions of the sizes of these perturbations are modeled. A test sample is similarly minimally perturbed by the detector from its decided-upon (target) class to every other (potential source) class. The p-values according to the corresponding null distributions of these test-sample perturbations are assessed using the corresponding nulls to decide whether the test sample is a TTE attack.

Systems, computer-implemented methods, and non-transitory machine-readable storage media are provided for detecting recapture attacks of images. One method comprises extracting one or more features from an image captured by a device; applying the one or more features as input to a trained machine learning model, wherein the trained machine learning model outputs a first score based on the extracted features; obtaining metadata of the image; performing a statistical analysis of the metadata of the image; generating a second score based on the statistical analysis of the metadata of the image; and generating a probability that the image is a recapture of an original image based on the first score and the second score.

This document describes techniques and apparatuses for verifying the authenticity of images. In aspects, methods include receiving, by a decoder system (220), an image (210) to be verified; performing feature recognition on the received image to determine determined features (238) of the received image; generating a first output (236) defining values representing the determined features; decoding the received image, by a message decoding neural network (252), to extract a signature (254) embedded in the received image, the embedded signature representing recovered features (258) of the received image; generating a second output (256) defining values representing the recovered features; providing the first output and the second output to a manipulation detection neural network (272); and generating, by the manipulation detection neural network, an estimation of an authenticity of the received image utilizing at least the first output and the second output.

A collation device includes a light source unit; a camera unit that receives light emitted from the light source unit and reflected in a collation area of an object to acquire a photographed image of the collation area; and a processor configured to, by executing a program: detect a positional relationship between the light source unit and the camera unit by using the photographed image, and notify of a collation result between the photographed image and a registered image prepared in advance by using the positional relationship.

The present application relates to a steganography method and a steganography apparatus using the same. According to the steganography method of the present application and the steganography apparatus using the same, they can simplify a data refinement process compared to the supervised learning method as AI learns a large amount of data related to steganography encoding and decoding by itself, can generate a high-quality stego video close to the cover video to improve the imperceptibility of hidden information as the generator and discriminator conduct mutual learning, can hide the message in the images and/or sounds according to the learning method to secure a high-capacity cache, and can secure robustness against third-party detection, monitoring, and removal attacks when learning the detection and avoidance methods of physical and technical detection systems (monitoring equipment, wiretapping equipment and security equipment, etc.). In addition, according to the steganography method of the present application, it can simplify a data refinement process compared to the supervised learning method as AI learns a large amount of data related to steganography decoding by itself, and can recover the hidden message close to the original hidden message to improve perceptibility of hidden information as the generator and the discriminator conduct mutual learning.