FixMyObjects

ML-based data checking tool for ICS2 and GDPR in linked data-trees of ONE Record

Comment

Inspiration

  • We want to explore how ONE Record and PayCargo provide a fruitful soil for an entire new ecosystem in the freight industry, providing mutual benefits for the supply chain and the new service providers.
  • In the legacy system, a lot of data transmitted through FWBs and FHLs contains dubious data in fields such as the Nature of Goods.
  • While ONE Record eases compliant data transmission massively, breaches of ICS2 and GDPR might still happen due to wrong contents in shipment, piece or product logistics objects. GDPR is especially relevant, as ONE Record allows to share pictures which might contain the faces of persons.
  • We tackle all of these issues at once and build an app for a dedicated checking company to find and eliminate ICS2- and GDPR-problems in linked ONE Record data and then charge the customer for the service performed.

Challenge #1: PayCargo challenge: We want to showcase how third party services could be offered on ONE Record. For all things payment, we use PayCargo to easily invoice customers on a per-service level.

Challenge #2: Safety & Security challenge: The service our app provides is, due to the challenges, linked to ICS2 and GDPR compliance, protecting personal data and ensuring that all pieces have valid information as required by customs.

What it does

  • FixMyObject is a lightweight, AI-based data checking tool handling whole trees of linked data to make sure no-one is processing ICS2- or GDPR-faulty data.
  • The backend GETs a tree of linked data from a live ONE Record server and flags all ICS2- and GDPR-issues in free text using Machine Learning.
  • The frontend has a data browser which allows o to review the detected issues, o flag additional things, o accept suggestions for ICS2 corrections and o anonymize all flagged personal data/blur faces of images.
  • Upon finishing the data check, the frontend transmits an invoice to the customer via PayCargo based on the number of checks and fixes and PATCHes the relevant objects via the ONE Record Server (not fully working due to problems with PATCH on embedded objects).
  • The app allows to scan an ONE Record NE:ONE label, take an image on of a damage, use AI to blur out faces of workers in the background (which are a GDPR breach), and PATCHing the data to the ONE Record Server (not fully working due to problems with PATCH on embedded objects).
  • The program is to be deployed with a third party of legal experts, providing data checking services to ONE Record users and invoicing them using PayCargo.
  • The app is to be used directly by shopfloor workers.

How we built it

  • The ONE Record Server is built with Java and is based on the IATA sandbox system
  • The frontend is built with React, visualising the shipment record (with all linked objects) in a data browser and highlighting potential ICS2 and GDPR issues and sending invoices via PayCargo.
  • The backend uses Google Data Loss Prevention to find ICS2 and GDPR issues with textual values in ONE Record and Google Vision for image segmentation and blurring of photos.
  • The native iOS and Android Apps are built with Flutter to take and send images connected to the relevant shipment to the backend for face finding and blurring.
  • A diagram of our architecture is attached.

Challenges we ran into

  • We were unable to PATCH back the results of the data check to ONE Record because our server implementation does not allow for the PATCH of embedded objects yet.

Accomplishments that we're proud of

  • Our solution is running cloud-native and container-based.
  • We combine state-of-the-art Machine Learning technology to provide a pre-compiled result to be double-checked by a human specialist.
  • We use first class payment solution to invoice customers for both ICS2- and GDPR-checks as well as corrections performed by specialists to the data.

What we learned

  • We learnt how to work with a running ONE Record server and performing tasks on the data

What's next for FixMyObjects

  • We have shown that Machine Learning can provide a feasible solution for handling ICS2- and GDPR-relevant issues with data shared via ONE Record, even if multiple companies are responsible for parts of the shipment record.
  • In the future, we hope to fix the bug to enable PATCHing embedded objects in our ONE Record server implementation. This allows Fix My Object to PATCH the changed data directly back to ONE Record.
  • Furthermore, we want to think about further services to be performed on ONE Record and invoiced with PayCargo to get a headstart on the exiting internet of logistics that is on the horizon.

Built With

Share this project:

Updates