Computer-implemented methods and systems are provided for digitally signing predetermined arrays of digital data. Such a method may provide a secret neural network model trained to classify arrays of digital data in dependence on data content of the arrays. The array of the arrays may be signed by supplying the array to the secret neural network model to obtain an initial classification result; and effecting a modification of data in the array to change the initial classification result to a predetermined, secret classification result, the modification being effected via a backpropagation process in the secret neural network model to progressively modify the array in response to backpropagated errors dependent on a difference between a current classification result for the array and the secret classification result.

An example symbol-based watermark detection method disclosed herein includes, in response to a comparison of a first count of occurrences of a first potential symbol value corresponding to a first symbol within a watermark and a second count of occurrences of a second potential symbol value corresponding to the first symbol, (i) determining a first accumulated signal to noise ratio value corresponding to the occurrences of the first potential symbol value, (ii) determining a second accumulated signal to noise ratio value corresponding to the occurrences of the second potential symbol value, and (iii) selecting one of the first or the second potential symbol value having a greatest accumulated signal to noise ratio value as a likely symbol value for the first symbol. The example method also includes concatenating the likely symbol value with other likely symbol values corresponding to other symbols of the watermark to detect the watermark.

A computer system for the creation of subliminal image or watermarks where the apparent video is an “obvious” image and a subliminal or “clandestine” image or watermark is hidden within the data structure. The concept is envisioned for the cloaking of images, sound, video or like digitized non-coded information. The cloaked files may be intended for storage, transmission, or clandestine placement in a public file system. The watermark may be used for the subliminal marking of a source of a file, its creation, or for tagging copyrighted information.

In an electronic device that embeds information in content, copyright protection and prevention of falsification are achieved at the same time. The electronic device includes a content generation unit, a first embedding processing unit, and a second embedding processing unit. In this electronic device, the content generation unit generates content. Furthermore, in the electronic device, the first embedding processing unit embeds, in the content, first embedded information related to the content. The second embedding processing unit embeds second embedded information in the content.

A method for checking the authenticity of products, by checking an image (A) of a product. The proof of authenticity is not visible to the human eye and cannot be copied. This is characterized in that a code stored in a halftone image by manipulation of dots and/or a manipulated field bounded in the halftone image can be read by means of an optical device and compared with a retrievable value in at least one database. In at least one field (F1 to F5) a part of a serial number is determined which describes the structure of the serial number and a hash function used for transmitting the serial number to the database, and this is also characterized in that the serial number is subsequently assembled and encrypted with the corresponding hash function.

This disclosure relates to advanced image signal processing technology including encoded signals and digital watermarking. The technology may be applied to retail packages and other printed objects, e.g., such as hang tags, labels and receipts.

Data processing is disclosed including acquiring a first target image to be embedded with information and to-be-embedded information, acquiring, using the first target image, a second target image corresponding to the first target image, the second target image corresponding to an image including low-luminance pixels in the first target image, the low-luminance pixels being pixels having a luminance no higher than a luminance threshold value in the first target image, selecting candidate image areas from the second target image, determining a target embedding position for the to-be-embedded information in the first target image based on the candidate image areas, and embedding the to-be-embedded information in the target embedding position in the first target image.

Provided is a computer-implemented video processing method. The method comprises receiving a stream of original images related to a video access event and creating a stream of output images corresponding to the original images. The output images include first images comprising a hidden digital forensic marker and second images comprising a visible digital forensic marker. The hidden marker and the visible marker each encode information related to the video access event. The stream of output images is output onto a network or caused to be displayed on a screen. The visible marker serves as a deterrent for distribution or recording, and if an attempt is made to remove it, the hidden marker remains in the image, allowing the information relevant to the video access event to be recovered. Also provided are a method of integrity detection for a stream of images containing markers, and a method of embedding a dynamic marker.

A method is provided for watermarking a machine learning model used for object detection. In the method, a first subset of a labeled set of ML training samples is selected. Each of one or more objects in the first subset includes a class label. A pixel pattern is selected to use as a watermark in the first subset of images. The pixel pattern is made partially transparent. A target class label is selected. One or more objects of the first subset of images are relabeled with the target class label. In another embodiment, the class labels are removed from objects in the subset of images instead of relabeling them. Each of the first subset of images is overlaid with the partially transparent and scaled pixel pattern. The ML model is trained with the set of training images and the first subset of images to produce a trained and watermarked ML model.

An information processing device, an information processing method, and an information processing program capable of reducing an information processing amount of an application processor that recognizes a subject from an input image are provided. An information processing device (100) according to the present disclosure includes an acquisition unit (4), a generation unit (7), and an output unit (6). From an imaging unit that captures an image and generates image data, the acquisition unit (4) acquires the image data. The generation unit (7) generates, from the image data acquired by the acquisition unit (4), metadata to assist an application processor that executes processing related to the image. The output unit (6) outputs the metadata generated by the generation unit (7) to the application processor.