With Deepfakes and Fake News on the rise, we fixated on solving the issue through decentralized provenance and cryptography. With NuCypher providing ways to achieve proxy re-encryption, an entity can encrypt content for a private group using time bound policies, we were inspired to utilize this technology to create a provenance chain that enables anyone to verify the origin and authenticity of published content. This in turn motivates organizations to publish highly sensitive and high impact contents like Earnings call/transcripts, C-Level Communiques, Town hall videos, etc through such a platform that can provide an intuitive method for verification resulting in increased confidence in published content.
What it does
TripleCheck now implements content publishing, encryption/decryption, and verification of published content. Users can login/signup using Fortmatic for easy on-boarding. Users can then create encryption policies for the content they produce. The publishers can then grant access to users in the platform with a set expiry date or publish directly to decentralized outlets like IPFS, Swarm. Platform members can verify hashes for privately shared content. The public can verify hashes for all TripleCheck enabled content that is published over the social medias, referenced in articles or news publications.
How we built it
We setup a Flask based API Backend that does much of the heavy lifting for processing data on the site. The frontend is NextJS based and it is served over a serverless architecture. We integrated Fortmatic to enable seamless user on-boarding. The site is capable of self-publication of private content and integrates IPFS for publishing to the decentralized web. A verify procedure essentially does "3 checks", OP's Key, in relation to the publishing Platform's Key and the related Decentralized link. Thus to enable decentralized proof and point of origin against fake news, if the content originates from TripleCheck, it is essentially authentic as long as the 3 checks tally up with the system.
Challenges we ran into
There were some challenges with combining Fortmatic and NextJS. We found early Sunday Morning that the latest nucypher 0.1.0a22 breaks a lot from the nucypher 0.1.0a21 that we initially started out with. Somewhere in the middle, we reset virtualenv and the new version broke our backend for several hours nearing submission.
Accomplishments that we're proud of
Having matured the idea from its initial concept of just protecting C-Level Communications to creating a decentralized universal provenance system is amazing. We were also pleasantly surprised to know that startups working to thwart deepfakes, fake news are on the RFS top 20 list published by Y Combinator.
What we learned
We quickly realized that using proxy-encryption on HD Videos is a very expensive task. We did theorize and validate that if we can sequentially pull frames every 5 secs and use that to render using the proxy re-encryption, it might be functional. We are still thinking on this one.
What's next for TripleCheck - Anti Deepfakes/Fake News through provenance
- Plans to utilize SigularityNet and determine plans for integration
- Plans to add more file types like excel sheets for Earnings Call, Quarterly Filing Reports, etc
- Plans to build the video proxy re-encryption engine may be utilizing GPU support