Abstract: Methods, systems, and devices for access management are described. A software platform may identify devices of a botnet based on a cluster score associated with a device characteristic. For example, the software platform may receive a request from a device to access an application. The software platform may determine a cluster score for the characteristic of the device. The cluster score may be based on a link between the device and a list of devices (e.g., devices of a botnet). If the cluster score satisfies (e.g., exceeds) a cluster score threshold, the software platform may deny the access request. In some examples, the cluster score may be determined using machine learning techniques. Based on determining the cluster score, the software platform may efficiently identify devices of the botnet and prevent brute force attacks, which may improve reliability of access for users of the application.

Abstract: An authentication system uses machine learning models to quantify a degree of risk that a given request to authenticate as a particular user of an organization is not in fact originating from that user, but rather from a malicious actor attempting to gain access to the user's account. More particularly, the authentication system employs both a user model that quantifies a degree of deviation from a user context in which a particular user typically requests authentication, and an organization model that quantifies a degree of deviation of a current context of the organization from a “normal” context for that organization. The user model and the organization can be employed individually, such as the organization model providing organization administrators with insights into the current security status of the organization, or together, such as using the risk scores of both models when assessing how to respond to a particular authentication request.

Abstract: A method authenticates users. During user enrollment, a computing device records 3D gesture samples captured by a depth sensor as performed by a first user. Each recorded gesture sample includes a temporal sequence of locations for multiple specified body parts. The device computes an average gesture, and selects an error tolerance. These are stored as a gesture template for the first user. A second user performs a gesture for authentication. The depth sensor captures a 3D gesture from the second user, where the captured 3D gesture includes a temporal sequence of locations for the multiple body parts. The device computes the distance between the captured 3D gesture and the average gesture. When the distance is less than the error tolerance, the second user is authenticated as the first user, and the device grants access to some secured features. Otherwise, the second user is not authenticated.

Abstract: Embodiments provide methods, apparatuses, and articles of manufacture for generating frames and changing lens positions based on the analysis of the frames. In one example, a controller coupled to a first camera and a second camera is to change a lens position of the first camera for a second frame based on the identification of a focal target in a frame generated by the second camera.

Abstract: A device to capture visual media, transiently store the visual media on a circular buffer, detect for a trigger from an environment around the device, and store the visual media on a location of a storage component separate from the circular buffer in response to detecting the trigger.

Abstract: Systems, methods, and machine-readable and executable instructions are provided for filtering image data. Filtering image data can include determining a desired depth of field of an image, determining a distance between a pixel of the image and the desired depth of field. Filtering image data can also include adjusting a contrast of the pixel in proportion to a magnitude of a weight of the pixel, wherein the weight is based on the distance.

Abstract: Embodiments provide methods, apparatuses, and articles of manufacture for generating frames and changing lens positions based on the analysis of the frames. In one example, a controller coupled to a first camera and a second camera is to change a lens position of the first camera for a second frame based on the identification of a focal target in a frame generated by the second camera.

Abstract: Implementations disclosed herein relate to determining an option based on a reaction to visual media content. In one implementation, a processor determines a user's degree of preference for visual media content and selects an option based on the determined degree of preference. The processor may provide the determined degree of preference.