Artificial Intelligence: The intelligent algorithm layer at the core of the anti-piracy framework comprises three key components:
• Anomaly detection
Building upon the Enhanced Customer 360 data set, we leveraged unsupervised auto-tuning anomaly detection algorithms to evaluate user behavior at the individual session/stream level. This approach enables continuous self-learning and adaptation, ensuring our system remains vigilant against evolving piracy trends without the need for manual intervention. Mathematically, the model can be summarized as follows:
R [0,100]
r (u, s) = f (Feature set 1, Feature set 2, Feature set 3)
where r (u, s) = risk score for user “u” during session “s”; f = anomaly detection model; Feature set 1–3 = features from Enhanced Customer 360 data set; R [0,100] = any real number between 0-100. While standalone anomaly detection plays a vital role, it is crucial to recognize that its effectiveness can be further enhanced through a holistic approach. Combining network analysis techniques and model explainability methods can significantly improve piracy identification’s precision while minimizing false positives.
• Network analysis
We constructed dynamic graphical networks that trace extensive syndicates of users, devices, and IP addresses linked to high-risk individuals identified by the anomaly detection framework. To illustrate the impact and application of this approach, consider these examples that offer a clear perspective on its effectiveness.
Example 1: Specific device mapped to identified pirates and at-risk non-pirates.
Example 2: Particular pirate mapped to multiple devices concurrently.
© 2023 Fractal Analytics Inc. All rights reserved
05
Made with FlippingBook - PDF hosting