Abstract: A method including: monitoring a set of streaming metrics for a video stream during a set of time intervals of a first duration during a first time window; in response to a first streaming metric, executing an image classification model based on a set of image frames in the video stream to characterize the image frames according to a set of tags; retrieving a content manifest associated with a content type of the video stream, the content manifest defining a set of target concepts related to the content type; deriving a difference between the set of tags to the set of target concepts in the content manifest to compute a match score for the video stream; in response to the match score exceeding a threshold score, flagging the video stream for manual authentication; and in response to receiving an abuse confirmation from the operator removing the video stream from the streaming platform.

Abstract: A method includes, at a first time: receiving a request for video content from a first user; generating a fingerprint for the first user; associating the first user with a first user population—assigned to a first CDN and receiving the video content from the first CDN during the first time period—based on the fingerprint; and accessing a first metric for distribution of video content from the first CDN to users of the first user population. The method also includes, at a second time: selecting a second user within the first user population; identifying a second CDN distinct from the first CDN; reassigning the second user to the second CDN; and accessing a second metric for distribution of the video content from the second CDN to the second user; and, in response to the second metric exceeding the first metric, reassigning the first user to the second CDN.

Abstract: A method including: monitoring a set of streaming metrics for a video stream during a set of time intervals of a first duration during a first time window; in response to a first streaming metric, executing an image classification model based on a set of image frames in the video stream to characterize the image frames according to a set of tags; retrieving a content manifest associated with a content type of the video stream, the content manifest defining a set of target concepts related to the content type; deriving a difference between the set of tags to the set of target concepts in the content manifest to compute a match score for the video stream; in response to the match score exceeding a threshold score, flagging the video stream for manual authentication; and in response to receiving an abuse confirmation from the operator removing the video stream from the streaming platform.

Abstract: A method includes, at a first time: receiving a request for video content from a first user; generating a fingerprint for the first user; associating the first user with a first user population—assigned to a first CDN and receiving the video content from the first CDN during the first time period—based on the fingerprint; and accessing a first metric for distribution of video content from the first CDN to users of the first user population. The method also includes, at a second time: selecting a second user within the first user population; identifying a second CDN distinct from the first CDN; reassigning the second user to the second CDN; and accessing a second metric for distribution of the video content from the second CDN to the second user; and, in response to the second metric exceeding the first metric, reassigning the first user to the second CDN.

Abstract: A method includes, at a first time: receiving a request for video content from a first user; generating a fingerprint for the first user; associating the first user with a first user population—assigned to a first CDN and receiving the video content from the first CDN during the first time period—based on the fingerprint; and accessing a first metric for distribution of video content from the first CDN to users of the first user population. The method also includes, at a second time: selecting a second user within the first user population; identifying a second CDN distinct from the first CDN; reassigning the second user to the second CDN; and accessing a second metric for distribution of the video content from the second CDN to the second user; and, in response to the second metric exceeding the first metric, reassigning the first user to the second CDN.

Abstract: A method including: monitoring a set of streaming metrics for a video stream during a set of time intervals of a first duration during a first time window; in response to a first streaming metric, executing an image classification model based on a set of image frames in the video stream to characterize the image frames according to a set of tags; retrieving a content manifest associated with a content type of the video stream, the content manifest defining a set of target concepts related to the content type; deriving a difference between the set of tags to the set of target concepts in the content manifest to compute a match score for the video stream; in response to the match score exceeding a threshold score, flagging the video stream for manual authentication; and in response to receiving an abuse confirmation from the operator removing the video stream from the streaming platform.