Streaming fraud detection using blockchain
Systems and methods for detecting fraudulent streaming activity. Streaming activity is posted to a blockchain by one or more DSPs. Blockchain streaming data is extracted from the blockchain and used as input in a machine learning model. The machine learning model takes the extracted blockchain data, along with additional inputs such as DSP trend pool and social pool inputs, and makes a determination regarding potentially fraudulent streaming activity.
1. A system for detecting fraudulent streaming activity, the system comprising:
a processor; and
memory, the memory storing instructions to cause a processor to execute a method, the method comprising:
extracting streaming data from a blockchain, the streaming data corresponding to play count activity for streaming media;
transforming the extracted streaming data into a format ingestible by a trained machine learning model;
inputting the transformed streaming data into the machine learning model, wherein the machine learning model comprises a trend pool that allows analysis of streaming patterns across multiple streaming platforms to identify discrepancies between the trends across platforms to suggest streaming fraud; and
determining that a particular stream is potentially fraudulent.
2. The system of claim 1 , wherein the machine learning model takes into account user engagement data including one or more of the following: gyroscope orientation, battery percentage, current city, track label, track distributor, and stream duration.
3. The system of claim 1 , wherein the extracted streaming data comprises hashes instead of raw data for zero knowledge proofs.
4. The system of claim 1 , wherein the machine learning model includes multiple machine learning algorithms for different attributes of streaming media.
5. The system of claim 1 , wherein the extracted streaming data corresponds to other information including one or more of the following: device battery information, operating system, user interaction data.
6. The system of claim 1 , wherein the machine learning model utilizes a social trend pool or a Digital Service Provider (DSP) trend pool as additional input.
7. The system of claim 1 , wherein the method further comprises posting the determination that the particular stream is potentially fraudulent to a data store for direct or indirect access by DSPs.
8. A method for detecting fraudulent streaming activity, the method comprising:
extracting streaming data from a blockchain, the streaming data corresponding to play count activity for streaming media;
transforming the extracted streaming data into a format ingestible by a trained machine learning model;
inputting the transformed streaming data into the machine learning model, wherein the machine learning model comprises a trend pool that allows analysis of streaming patterns across multiple streaming platforms to identify discrepancies between the trends across platforms to suggest streaming fraud; and
determining that a particular stream is potentially fraudulent.
9. The method of claim 8 , wherein the machine learning model takes into account user engagement data including one or more of the following: gyroscope orientation, battery percentage, current city, track label, track distributor, and stream duration.
10. The method of claim 8 , wherein the extracted streaming data comprises hashes instead of raw data for zero knowledge proofs.
11. The method of claim 8 , wherein the machine learning model includes multiple machine learning algorithms for different attributes of streaming media.
12. The method of claim 8 , wherein the extracted streaming data corresponds to other information including one or more of the following: device battery information, operating system, user interaction data.
13. The method of claim 8 , wherein the machine learning model utilizes a social trend pool or a DSP trend pool as additional input.
14. The method of claim 8 , further comprising posting the determination that the particular stream is potentially fraudulent to a data store for direct or indirect access by DSPs.
15. A non-transitory computer readable medium storing instructions to execute a method, the method comprising:
extracting streaming data from a blockchain, the streaming data corresponding to play count activity for streaming media;
transforming the extracted streaming data into a format ingestible by a trained machine learning model;
inputting the transformed streaming data into the machine learning model, wherein the machine learning model comprises a trend pool that allows analysis of streaming patterns across multiple streaming platforms to identify discrepancies between the trends across platforms to suggest streaming fraud; and
determining that a particular stream is potentially fraudulent.
16. The non-transitory computer readable medium of claim 15 , wherein the machine learning model takes into account user engagement data including one or more of the following: gyroscope orientation, battery percentage, current city, track label, track distributor, and stream duration.
17. The non-transitory computer readable medium of claim 15 , wherein the extracted streaming data comprises hashes instead of raw data for zero knowledge proofs.
18. The non-transitory computer readable medium of claim 15 , wherein the machine learning model includes multiple machine learning algorithms for different attributes of streaming media.
19. The non-transitory computer readable medium of claim 15 , wherein the extracted streaming data corresponds to other information including one or more of the following: device battery information, operating system, user interaction data.
20. The non-transitory computer readable medium of claim 15 , wherein the machine learning model utilizes a social trend pool or a DSP trend pool as additional input.