Endangered Privacy: Large-Scale Monitoring of Video Streaming Services
summary
A protocol-agnostic monitoring system is built that collects manifest-based fingerprints, stores sliding windows of segment sizes in a k-d tree and uses a TShark wrapper plus time-series heuristics to identify streams in real time. It processes a dataset of 240,000 videos and 3 million fingerprints and achieves over 99.5% accuracy after minutes of eavesdropping, works even behind VPNs or passive Wi‑Fi sniffing, and shows traffic padding mitigations cost ≈2.2× bandwidth while buffer-management fixes are proposed.
AI-written summary · about
chapters (14)
- 0:00Video streaming privacy risks
- 0:34Adaptive bitrate streaming
- 2:17Network bursts reveal content
- 3:00Previous stream monitoring research
- 4:04Goal of large-scale monitoring
- 4:45Attacker models: strong vs weak
- 6:12Fingerprint data collection method
- 7:38Live attack with network bursts
- 8:43Identification heuristic and scoring
- 9:25Real-world evaluation
- 10:30Accuracy and false positives
- 11:57Base rate challenges in VPN/Wi-Fi
- 12:17Mitigation strategies and protocol leaks
- 13:04Conclusion
related talks
13mSecurity Theater: Moving Beyond Performative Content Protection
08mThe do's and don'ts about Streaming security...
17mPrivateEye: Scalable and Privacy Preserving Compromise Detection in the Cloud
29mYou Can’t Hide: How a Device Can Give Away a Fraudster’s Secrets.
30mSpying on my Network for a Day: Data Analysis for Networks
22mEffective Detection of Multimedia Protocol Tunneling using Machine Learning
27mCollect All the Data; Protect All the Things
22mProbabilistic Spying on Encrypted Tunnels