Encrypted Media Identification via Traffic Fingerprinting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying pre-recorded media content in encrypted networks face challenges due to network noise, retransmissions, and congestion, making it difficult to scale for real-time processing and accurate analytics, especially in packet data networks.
Innovation Solution
A system that separates HTTPS traffic into audio and video segments, calculates segment sizes, maintains a sliding window of segment sizes to form a fingerprint, and matches the fingerprint with a reference catalog using a k-d tree key calculation, enabling identification of media content despite encryption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are used to identify media content in encrypted networks, then identification accuracy can be improved in controlled environments, but the approach becomes non-scalable and model accuracy decreases significantly with scale
Solution Approach 1:
The patent segments the identification process into distinct components: extracting network characteristics from encrypted traffic, creating fingerprints from these characteristics, and matching fingerprints against a database. This segmentation allows the system to scale by processing traffic flows independently and maintaining a centralized fingerprint database, avoiding the scalability limitations of monolithic machine learning models.
2Reliability
If network encryption is implemented to protect PII data, then data security is improved, but access to usage data is diminished making network analytics unobtainable
Solution Approach 1:
The patent introduces an intermediary approach by extracting metadata characteristics (packet sizes, timing intervals, flow patterns) from encrypted traffic without decrypting the content. These characteristics serve as intermediaries that preserve privacy while enabling analytics, allowing the system to gain insights into usage patterns without compromising data security.
Solution Approach 2:
The patent replaces traditional decryption-based analytics (mechanical system requiring key access) with characteristic-based fingerprinting. Instead of mechanically decrypting and analyzing content, the system substitutes this with extracting and comparing network-layer characteristics, achieving analytics without breaking encryption.
3Measurement precision
If fingerprint matching with k-d tree is used to identify media content, then real-time identification accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-processing network traffic to extract characteristics and generate fingerprints before matching. The k-d tree database is pre-built and organized in advance, allowing efficient real-time matching without complex runtime processing. This preliminary preparation reduces the complexity of the matching operation itself.
Data Source
AI summary
Aspects of the subject disclosure may include, for example, a device having a processing system including a processor; and a memory that stores executable instructions that, when executed by the processing system, facilitate performance of operations including receiving encrypted hypertext transport protocol (HTTPS) traffic including media content; separating the HTTPS traffic into audio segments and video segments; calculating a size for each audio segment in the HTTPS traffic; maintaining a sliding window of a plurality of sizes of consecutive audio segments to form a fingerprint; and identifying the media content by matching the fingerprint with a reference in a catalog. Other embodiments are disclosed.