Inferring ABR Video Streaming Behavior from Encrypted Traffic

The patent introduces an intermediary inference system that acts as a mediator between encrypted traffic and analysis requirements. Instead of directly analyzing encrypted traffic (which loses information), the system uses observable traffic characteristics (packet sizes, timing patterns, sequence numbers) as intermediaries to infer ABR streaming behavior, segment boundaries, and quality metrics without decrypting the content.

The patent replaces the traditional mechanical approach of direct traffic inspection (which requires unencrypted data) with an inference-based system that uses statistical patterns and machine learning. The system substitutes direct observation with indirect reasoning, using observable traffic patterns to deduce unobservable streaming parameters through probabilistic models.

The patent changes the parameters being measured from content-based metrics (which require decryption) to traffic-pattern-based metrics (observable in encrypted traffic). The system measures packet size distributions, inter-packet timing, sequence number increments, and flow patterns as alternative parameters that preserve client behavior information while maintaining encryption.

The patent implements feedback loops where inferred behavior patterns are continuously refined against new traffic observations. The system uses feedback from traffic pattern matching and segment sequence inference to improve its understanding of client behavior over time, even though the underlying traffic remains encrypted.