Abstract: System and method of performing identity authentication based on de-identified data, with receiving of at least one data item from a user, creating at least one de-identified data item and a corresponding authentication key, based on the received at least one data item, storing the at least one de-identified data item in a database, receiving at least one new data item for authentication, retrieving the authentication key corresponding to the user, generating at least one new de-identified data item based on the received at least one new data item, and comparing the at least one new de-identified data item with the stored at least one de-identified data item to authenticate the identity of the user.

Abstract: System and method for training a human perception predictor to determine level of perceived similarity between data samples, the method including: receiving at least one media file, determining at least one identification region for each media file, applying at least one transformation on each identification region for each media file until at least one modified media file is created, receiving input regarding similarity between each modified media file and the corresponding received media file, and training a machine learning model with an objective function configured to predict similarity between media files by a human observer in accordance with the received input.

Abstract: A system and a method of prevention and/or detection of adversarial attacks against a classifier may include for example: using a classifier adapted to classify data elements of a specific category; receiving a first data element; applying the classifier on the first data element, to produce a first classification of the data element; using an autoencoder to generate a reconstructed, second data element based on the first data element; applying the classifier on the second data element, to produce a second classification of the data element; and analyzing the first and second classifications to detect and/or prevent an adversarial attack on the classifier.

Abstract: A system and a method of performing face recognition may include: receiving a first facial image, depicting a first face, and a second facial image depicting a second face; applying an ML model on the first image, to produce a first representation vector, and applying the ML model on the second image to produce a second representation vector; comparing the first representation vector and the second representation vector; and associating the first face with the second face based on the comparison, where the ML model is trained to produce the representation vectors from the facial images, based on regions in the facial images that correspond to distinctiveness scores that are beneath a distinctiveness threshold.

Abstract: A system and method for concealing and revealing a human face in media objects may include obtaining a first media object capturing an image of the face; employing a first unit to: extract, from the first media object, a set of features representing the face, and generate a second media object, by embedding the extracted features in an anonymized media object; and employing a second unit to recognize the face based on the second media object.

Abstract: To consistently anonymize video, an identity-preserving encoder may be trained with a decoder to preserve identity information unique to an individual in an encoded space. The identity-preserving encoder may encode source video images of a source identity to generate identity-preserving encoded data representing the source identity. A de-identification engine may generate de-identifying encoded data representing a different target identity. A transformation may be generated based on a difference between the identity-preserving and de-identifying encoded data, representing swapping the source identity with the target identity. The identity may be consistently swapped from the source to target in the encoded space by applying the same transformation to the identity-preserving encoded data for the source images.

Abstract: A system and method of anonymization of a face in a set of images by at least one processor, the method including: receiving a first set of images; extracting from the first set of images a first face, depicting a person and having a first set of attributes; and performing a perturbation of the first face to produce a second face having a second set of attributes, wherein the second set of attributes is adapted to be visually perceived by a viewing entity as being substantially equivalent to the first set of attributes, and wherein the second face is adapted to be visually perceived by a viewing entity as not pertaining to the depicted person.

Abstract: A system and method of performing facial image anonymization may include: obtaining, by at least one processor of a first computing device, at least one first image data element, depicting a first face; using, by the at least one processor, a first neural network (NN), to encode the facial image into a first low-dimension, latent vector; using, by the at least one processor, a transformation module, adapted to modify the first latent vector, to produce an anonymized latent vector; and transmitting, by the at least one processor, the anonymized latent vector, via a communication network to a remote second computing device, wherein the second computing device is adapted to decode the anonymized latent vector into a second image data element, depicting an anonymized face.

Abstract: To consistently anonymize video, an identity-preserving encoder may be trained with a decoder to preserve identity information unique to an individual in an encoded space. The identity-preserving encoder may encode source video images of a source identity to generate identity-preserving encoded data representing the source identity. A de-identification engine may generate de-identifying encoded data representing a different target identity. A transformation may be generated based on a difference between the identity-preserving and de-identifying encoded data, representing swapping the source identity with the target identity. The identity may be consistently swapped from the source to target in the encoded space by applying the same transformation to the identity-preserving encoded data for the source images.

Abstract: System and method of performing identity authentication based on de-identified data, with receiving of at least one data item from a user, creating at least one de-identified data item and a corresponding authentication key, based on the received at least one data item, storing the at least one de-identified data item in a database, receiving at least one new data item for authentication, retrieving the authentication key corresponding to the user, generating at least one new de-identified data item based on the received at least one new data item, and comparing the at least one new de-identified data item with the stored at least one de-identified data item to authenticate the identity of the user.

Abstract: Non-rigid dense correspondence (NRDC) for image enhancement may be performed. In one embodiment, a correspondence may be computed for each one of a plurality of regions of a source image to one of a plurality of regions of a reference image. Computing the correspondences may include searching within a search range for each of a plurality of image characteristics. One or more of the correspondences may be aggregated into matched regions. A global color transform and/or deblurring may be applied to the source image. In one embodiment, the global color transform and/or deblurring may be based on the matched regions. At least one of the search ranges may optionally be adjusted. In some embodiments, computing, aggregating, applying and/or deblurring, and adjusting may be iteratively performed.

Abstract: Non-rigid dense correspondence (NRDC) for image enhancement may be performed. In one embodiment, a correspondence may be computed for each one of a plurality of regions of a source image to one of a plurality of regions of a reference image. Computing the correspondences may include searching within a search range for each of a plurality of image characteristics. One or more of the correspondences may be aggregated into matched regions. A global color transform and/or deblurring may be applied to the source image. In one embodiment, the global color transform and/or deblurring may be based on the matched regions. At least one of the search ranges may optionally be adjusted. In some embodiments, computing, aggregating, applying and/or deblurring, and adjusting may be iteratively performed.