Empowering Sexual Assault Survivors Through Cryptography | Samantha Ouyang - Zuvivor, FranklinDAO

Transcript

[Music] all right hello everyone my name is Samantha oang I'm a 20-year-old Nomad on a gap Year from hupen MNT right now and I've been working in crypto since I was 16 um and so starting from last October where I spent two months in Thailand at multiple different pop-up cities I've been working on something called zuvor which is a platform that empowers sexual assault survivors to report their abusers by matching victims of the same perpetrator together so that they know that they're not alone and can find a path forward together so a little bit about my personal story so I've been traveling around for crypto conferences since I was 17 and the first time that ever I

was ever sexually harassed was when I was in Austin for South by Southwest that year um and that essentially made me hit the city ever since um but I ended up returning to Austin two years later for a consensus so this was um around May last year and I started working on this idea in honor of my younger self um and at megaz an edge city um in Thailand I became re-inspired to build this project out uh because I actually had to report my Airbnb host for sexual assault in Cambodia right before coming to changai and I myself was so scared to report and the numbers reflect this too 75% of sexual assault cases go unreported victims are afraid to speak up because of fear of retaliation

defamation lawsuits reputation loss and all of these other reasons listed here that create something called first mover disadvantage and so Zu Viber Works to eliminate that first mover disadvantage by allowing survivors to create a Time stamped record of their assault um and be matched with victims of the same perpetrator victims can submit a report where they can fill as many identifiers as they know about the perpetrator including things like name email socials uh physical features location Etc document what happened and upload evidence and then verify their Identity or kyc using ZK email or World ID um and this is essentially to prevent false reports or defamation um and then I also previously successfully implemented uh GPT 4 when it first came out uh to be

able to describe um an uploaded image and that would basically be able to help process image data submitted in reports um and then second building on top of cursives technology the system leverages multiparty computation or MPC specific uh specifically privates the intersection to identify overlapping information um in the rep report submitted without revealing anything else about uh their identity um and then from here a network of victim nodes developed from overlapping data in the edges and so this forms a knowledge a Knowledge Graph structure um and so clusters will eventually start to form around specific perpetrators identified as more pieces of the puzzle are revealed from different victim stories to form a more comprehensive story um and the Survivor nodes May gain access to the network of other survivors in the cluster um and then lastly once victims

become part of a cluster they can check their perpetrators to see the number of reports filed against them um and match with other victims of the same offender um the platform also noted notifies victims when these matches occur and offers them the option to connect in an anonymous private chat to find support and solidarity with each other um and so this is just like a little bit of a demo of the um implementing GPT 40 and essentially it's able to identify the uh things in an image um and then translate that to text um here's just a little bit of a demo of the front end so essentially uh you fill out a form uh to essentially report a perpetrator and then all of the fields are optional to fill out um and

you essentially want to fill out as many things that you know about uh the person um and then at the end you choose a verification method so here I'm using World ID to essentially verify that I am a human submitting this report um and then after that is done uh you can go back to the original it brings you back to the original page and you click submit report um and that's processed on chain um I also use arbitrum stylist to help me write my smart contract and this was pretty much used to store data about the reports instructs and right related functions as well as help identify and count the overlapping reports um and then after Devcon so at zo Thailand I started working more on the graph architecture that I mentioned

earlier and so this was a little bit of a whiteboarding session that I did uh while um a while back while I was still teasing out the details and so first I worked on training a graph neural network using pytorch for the matching algorithm and so the different fields like ey color license plate or phone number are encoded or embedded to become node features and then uh the GNN or graph no network learns these embeddings so that it can identify or classify um the whether these nodes represent the same perpetrator or not um and then now with it GNN and then using Ezekiel um very funny I actually bumped into the CEO of Ezekiel this morning when I went to torch's Tacos um it allows you to prove the trustworthiness of the model um and that it's acting in good faith without having to reveal any of the

sensitive data submitted by the victims um and so now I'm just going go into Quick uh walk through with some sample data so um this is the information that essentially is collected through the form and as you can see here even though these two reports aren't exactly the same with the same strings uh you can tell they're probably talking about the same uh the same perpetrator given that it's talking about a John Doe and a John D um and they're both uh one is Hispanic and one is Mexican and it's happening a similar area um and so there it brings a higher confidence rate um and then you can probably tell that they're likely the same person that puts it into the same cluster um and then here's also a similar example except this is for Jane Smith um you can see that she's white um and one

is Caucasian and then they're both located in NYC or New York um and then lastly here is a little bit of an outlier so report 4 has um no commonalities compared to any of the previous ones and so you end up with like a graph structure that looks sort of like this um where the ones that are similar are grouped together um and the ones that are different are grouped in a different one um and so here's just uh me running the program and so essentially it creates vectors of different um node feature en codings and at the bottom it generates uh edges if they are similar enough um and so these are bidirectional edges and So based on the graph before uh you would see that zero and one are linked and then one and zero are also linked

and then 2 three and 32 are linked um and if you line up the array on the edge indexes on the side you can see that um it is the case and you also see that four is no to be found because it does have no edges attached to it um here's just a little bit of like the Tex stack that I used to build this um and yeah here's some of the traction that I've gotten so far so uh since starting this project in October um it's gotten a lot of traction that I'm really grateful for um it's gotten a lot of widespread support from people who are really eager to bring it back to their own respective communities um like I've gotten people who are interested from like the modeling communities Jiu-Jitsu um even like the club scene in Berlin and so many anym um and I started

working on this around the time of the news about DWF Labs um and then from there I shared my own personal story about what happened in Cambodia and also my wife were working on the project um and then that tweet sort of went semi viral as the crypto Community really came together uh in support of Sexual Assault Awareness and safety at networking events um and my story was seen by a lot of different notable figures that I've been following for a long time um like Laura Shin like the really famous podcast host and also the executive producer of the metallic movie um and then at Edge City I ended up uh winning four prizes at the hackathon as well including first place in the arbit track and also the Community Choice prize which was essentially where they just set every finalist uh project on stage um and whichever got the loudest

cheer ended up winning and it just felt really heartwarming to have an entire crowd just cheering at the top of their lungs for me um and then after that the judge from arbitrum actually did an interview with me um and that was posted on Twitter it got reped posted and quote tweeted a lot um felt a lot of support from there um and after that people just began sending me money to my evm address and so that was really crazy um and so on the bottom right I think you can actually see I like told my dad about oh like someone gave me money for this and he was like oh yeah uh he was basically Jo just like joking about how like I shouldn't disappoint my investors um and yeah after Thailand um I'm just following up with a lot of the

ecosystems that were really interested in supporting um and just looking into different grants and I'm also looking for someone uh like a technical co-founder or anyone who would want to help on this project um and so yeah thank you so much for your time it's been an absolute pleasure to speak at eth Denver this year um and let's come uh together to hold perpetrators accountable thank [Applause] you for