Abstract: Techniques for targeting media content items for weakly connected devices are described. In some embodiments, a model building method in preparation for approximated personalization is performed at a server. The server uses data from fully connected devices to build a full model, which classifies profiles into clusters and tags targeted content with content rankings. The server also builds device models that approximate the full model. The server then sends the device models to user devices for targeted content selection. In some embodiments, a content selection method for approximated personalization is performed at a user device. The user device receives parameters of a device model trained by a server. The user device then executes the device model using local user data to assign a profile to a cluster and selects targeted content for the profile based on the cluster assignment and the rankings of the targeted content.

Abstract: Techniques for a trusted system for secure content distribution and trusted recording of content consumption are described. In some embodiments, the trusted system transcodes and transcrypts a media content item using a key obtained from a content provider and one or more keys based on an entitlement from a service provider to generate an encrypted media content item. The trusted system further receives a request to provide the media content item to a client device. The trusted system also obtains a signed audit token recording the request upon an authorization by the service provider based on the entitlement and a confirmation by the content provider, where the signed audit token is signed by the content provider and the service provider. The trusted system additionally provides the one or more keys for decrypting the encrypted media content item and reports the signed audit token.

Abstract: Techniques for targeting media content items for weakly connected devices are described. In some embodiments, a model building method in preparation for approximated personalization is performed at a server. The server uses data from fully connected devices to build a full model, which classifies profiles into clusters and tags targeted content with content rankings. The server also builds device models that approximate the full model. The server then sends the device models to user devices for targeted content selection. In some embodiments, a content selection method for approximated personalization is performed at a user device. The user device receives parameters of a device model trained by a server. The user device then executes the device model using local user data to assign a profile to a cluster and selects targeted content for the profile based on the cluster assignment and the rankings of the targeted content.

Abstract: Techniques for targeting media content items for weakly connected devices are described. In some embodiments, a model building method in preparation for approximated personalization is performed at a server. The server uses data from fully connected devices to build a full model, which classifies profiles into clusters and tags targeted content with content rankings. The server also builds device models that approximate the full model. The server then sends the device models to user devices for targeted content selection. In some embodiments, a content selection method for approximated personalization is performed at a user device. The user device receives parameters of a device model trained by a server. The user device then executes the device model using local user data to assign a profile to a cluster and selects targeted content for the profile based on the cluster assignment and the rankings of the targeted content.

Abstract: Techniques for targeting media content items for weakly connected devices are described. In some embodiments, a model building method in preparation for approximated personalization is performed at a server. The server uses data from fully connected devices to build a full model, which classifies profiles into clusters and tags targeted content with content rankings. The server also builds device models that approximate the full model. The server then sends the device models to user devices for targeted content selection. In some embodiments, a content selection method for approximated personalization is performed at a user device. The user device receives parameters of a device model trained by a server. The user device then executes the device model using local user data to assign a profile to a cluster and selects targeted content for the profile based on the cluster assignment and the rankings of the targeted content.

Abstract: A method for managing content in a network having one or more user devices and one or more edge computing systems includes determining a likelihood that a user device will connect to an edge computing system during a time window, based on the determined likelihood, storing on the edge computing system information reflecting an activity pattern associated with the user device, receiving a request for content from the user device, generating metadata based on the stored information and the request, and transmitting the metadata to the user device.

Abstract: A method for managing content in a network having one or more user devices and one or more edge computing systems includes determining a likelihood that a user device will connect to an edge computing system during a time window, based on the determined likelihood, storing on the edge computing system information reflecting an activity pattern associated with the user device, receiving a request for content from the user device, generating metadata based on the stored information and the request, and transmitting the metadata to the user device.

Abstract: Server-side trick mode restrictions in a stream may be provided. The stream may comprise, for example, a Video on Demand (VOD) Adaptive Bitrate (ABR) stream. First, a playlist request corresponding to a stream may be received from a user device. Next, a current stream position in the stream may be determined. In response to receiving the playlist request and when the current stream position corresponds to a non-skip segment, a playlist may be created having a last chunk comprising a next chunk of the stream from the current stream position. The playlist may then be provided to the user device.

Abstract: Server-side trick mode restrictions in a stream may be provided. The stream may comprise, for example, a Video on Demand (VOD) Adaptive Bitrate (ABR) stream. First, a playlist request corresponding to a stream may be received from a user device. Next, a current stream position in the stream may be determined. In response to receiving the playlist request and when the current stream position corresponds to a non-skip segment, a playlist may be created having a last chunk comprising a next chunk of the stream from the current stream position. The playlist may then be provided to the user device.