Topics

Tracking harmful content on a blockchain


Middleton, Dan
 

I saw this press release about facebook open sourcing video pattern matching which “it uses to identify child sexual exploitation, terrorist propaganda, and graphic violence”

https://www.theverge.com/2019/8/1/20750752/facebook-child-exploitation-terrorism-open-source-algorithm-pdq-tmk

 

One aspect of the code is fingerprinting media. Those hashes are stored in files about ¼ MB. That’s a little big for typical blockchain designs, but not out of the question and also a good requirement to feed into the projects.

 

https://github.com/facebook/ThreatExchange/tree/master/hashing/tmk/sample-hashes

 

I know little about this space beyond what the articles mention. It appears that there are independent databases that people cross reference. I’m wondering if there would be a benefit to unifying those datasets on a single blockchain. Off the top of my head it seems to match some basic blockchain criteria… you have independent actors that all need to read/write to the same database, there’s a good reason for the data to be public, and there’s resiliency benefits to replicating the data.

 

Users (e.g. social networks, media companies) retaining local replicas via their blockchain instances would also benefit from data locality in running their queries.

 

Regards,

 

Dan Middleton

Principal Engineer

Intel

 


Karen Ottoni
 

Thank you for sharing Dan, this is an interesting idea, bringing those siloed databases could help identify bad actors more easily or quickly.

>>Brian Iselin of Slavefreetrade.org would probably have good input here on what exists, works, and doesn't work and I know he's been working away on something related to this topic the last year....Brian do you have anything to add? Anyone else?

Dan/Brian: Please feel free to bring up this discussion point on a future call!