- Justas Cepenas
- Tautvydas Versockas
- Laura Silvanavičiūtė
- Mindaugas J
- Inga Lasauske
- Simona Kaveckaite
- Laimonas Miseikis
- Nerijus Lasauskas
In these difficult times we already have a lot of tools that provides ever growing charts, birds eye view maps with red dots/locations, etc. to cover situation outside. While these tools are useful for professionals, for ordinary people it says not that much. He or she have much more personal questions about the situation we are in, like "What about my environment ?", "Since this outbreak started, places I visited were safe?" etc. These are important questions to answer and to do so, it is possible only with personalized data.
Our teams' main goal was to investigate if we can answer these questions with data that already exists and accurate enough to make informative recommendations. To solve this problem there is a need to have active/passive tracking system to tackle case by case COVID-19 location cases without impacting too much of peoples day to day activities and privacy. What if it would be possible to do so without installing any additional apps and having accurate location service to check on. And apparently it is! With Google Timeline:
Google Maps Timeline service have enormous data set of places where people been on any given time frame. If such big data set could be used to identify COVID-19 cases and their whereabouts in past weeks, it could be possible to build service that checks if anyone been in potentially contaminated places. Based on that non medical recommendations could be given and directions to local health authorities provided.
Its very simple to use our system: no additional apps to install (if you already have Google Maps), just make sure you "Location History" is enabled. After this is done, download PVG chrome extension which will be able to check your location against our database of potentially infected people. If you already diagnosed as COVID-19 positive, you could serve good cause and upload your location history of past three weeks. NO identifiers required and all privacy aspects are respected.
You can download fully working Chrome Extension from here: Privacy VS Germs Extension
How we built it (+team)
Idea was pitched among friends on 2020-03-26. No surprise that a bunch of good old friends volunteered to take a part in it. We started at Swiss based hackaton codeVScovid19 where we built basic concept which formed main building block, in Hack The Crisis Denmark we have introduced new features and in The Global Hack we have worked on email privacy cases and tried to solve data issues.
Guy who keeps mood and enthusiasms alive:Justas Cepenas - initial idea
Data Science Team:
- Simona Kaveckaite - data models and PowerBI
- Tautvydas Versockas - back-end hacker
- Mindaugas Jurgelaitis - back-end guru
- Laura Silvanaviciute - front-end super-star (and you should definitely check her performance: https://youtu.be/hqWIdaAjdmI?t=1985 How cool is that?)
- Inga Lasauske - creative lead
Tech. stack was on purpose chosen to have good compromise of rapid development and scalability in mind, hence DB is ElasticSearch, backend - node.js, frontend(chrome extension) - Angular 9, number crunching - PowerBI, hosting in AWS cloud.
As our main data source we choose Google Maps Timeline. This is the backbone of our system. Data is already there, all we had to do it is to take it, present it nicely to user and transform it to make this data alive.
Though out the global hack hackaton:
- Backend validations and data correctness
- Privacy oriented workflows. Email based management without storing any sensitive data: Gliffy Diagram
- User experience changes and additional features
- Data model enhancements. Discussions with mentors about AI applications
- User interface survey conducted
- Begin to actively contact positive cases by phone
- System is able to access approximated locations
- Email authorization system developed
- Data model was expanded to facilitate aprox. Locations
- Explanatory video and pdf guidelines were created
Challenges we ran into
Google Maps does not provide any API or access points to users Timeline (Location History) data. When you look up and see our teams' description, there are no surprises that we found a way how to extract this data in systematic way and most important respecting all privacy requirements.
As have created our main goal as "privacy first" we had a lot of discussions how to full preserve it. In attached files you will see proposed architecture of storing locations data without having any identifiers that can lead to locations stored.
Accomplishments that we are proud of
We managed to build quite complex data models that are based on feedback given by medical professionals. And to wrap all technicalities we have beautiful visuals for you to enjoy. Improved privacy properties was this weeks highlight which makes this POC even more closer to end product.
What we learned
With a good team quarantine weekend can be fun and very productive!
What's next for Privacy VS Germs
Just a simple fact that we can have data base of 3 weeks whereabouts of infected people opens up a lot of thoughts for future development. just few to mention:
- We built our system with Open API in mind, that means that and service can provide us set locations and we will be able to tell if given data set collide with our records.
- As database for infected peoples' location history increases such resource can be valuable for very complex virus spread modeling. We could potentially identify source clusters, infection spreading patterns, etc.
- As we do have history of every movement, we could identify travels from abroad and other more advanced location change patterns.
- Advanced data processing with AI systems.
Try It out
angular.js, elasticsearch, node.js, powerbi