Automotive data task force Teleconference

<scribe> scribenick: ted

<scribe> scribe: Ted

SPARQL endpoint for VSS

http://automotive.eurecom.fr/simulator/query/

Benjamin: this is a prototype of providing vehicle data via a sparql endpoint
... it is a demonstration of a semantic data access would look like

<uk> hi Ted, our WebEx conference link doesn't seem to work. Could you please re-send the invite? Thanks

Glenn: these the types of queries developers would construct?

Benjamin: the queries present attributes, signals when they have them
... you can make various types of queries such as which windows are currently open

Ted: thank you Benjamin, PoC always helpful for people to get their heads around something. this is geared more towards a single vehicle at a specific point in time whereas we will be focusing on aggregate datasets from multiple vehicles

Glenn: we may look at a range of data, isolated to a single geographic area for instance

Benjamin: this is just a small sample/example and sparql can handle multiple vehicles and larger sets of data
... there is no limitation on the types of queries you can do
... you would be able to put in T1 and T2 or geographic range
... this is one of the really interesting aspects of sparql

Ted: perhaps Geotab, Caruso or BMW can provide an anonymized dataset you can use

Glenn: Geotab has recently gone live with an anonymized public data set https://data.geotab.com/

Benjamin: I have another project coming up on tragectory data mining using vehicle data
... using just a few signals. I should be able to get a dataset of multiple cars
... not sure that is of interest to this group

Ted: definitely interested @@examples

continuing Capturing consent

Ted: we ran out of time for questions and comments. basically I want people to have a firmer understanding on the solution and document it so we can share widely for input

Ulrich: want to hear questions people didn't have time for last week
... two slide decks were shared with the group, first was what was presented last week, second more details with granular interactions

Glenn: first thank you for sharing your consent model
... it is highly applicable for an open standard
... on slide 4 there was a discussion on balancing on interest
... it should have the broadest governance identified. is this EU centric or more global?

Ulrich: it was taken directly from GDPR and afraid I cannot speak to applicability in other regions, I can ask our legal team and get back to you
... perhaps others have experience
... as you infer it is of interest to share specific user information to a third party

Glenn: Geotab is in exclusively in commercial fleet model. in our case the driver is not the owner of the vehicle
... for our consent model we use and EULA with the fleet manager through a click through
... we defer to the fleet owner to get consent from the underlying drivers

Ulrich: we have started to consider that, see slide #25
... we want to include it but it is not part of the current model
... we are discussing this with a particular OEM
... they have two paths for handling, one is contract based consent as you describe and other for personal vehicles

Dominik: current assumption is model we have for private vehicles could be applicable to fleet case and we are currently working on it
... contract approach is done in the real world at present

Glenn: I could bring in our chief privacy officer for a subsequent call

Ulrich: our goal is to have one infrastructure that can handle the multiple use cases
... technically, ideally one protocol can cover all those cases

Tim: I have a more technical question
... reliance on a neutral server may be a weak point. has there been any exploration at looking at decentralized approach eg blockchain?

Ulrich: we started with presumption of authority server to be centralized and we have not looked at decentralized yet

Dominik: we have a footnote mentioning how this could be done with blockchain
... the token management is handled by neutral server. there could be disadvantages to going decentralized
... for example you might have a situation where data is shared along different layers and competitors could deduce data activity based on decentralized information
... we will be investigating it further and enumerate advantages and disadvantages

Tim: as you think about some of the privacy requirements you might want to look at moneros ring encryption model
... it is more costly than etherium blockchain but provides transactional blindness and yield what you are looking for
... the volume of data (auth tokens) is not substantive enough it could be fine

Dominik: agree there could be variants that provide the protections we are looking for

Ulrich: we would appreciate any pointers you can share with us and even if you are willing to explore it with us
... trusting one authorization provider is certainly a problem
... please go through the second slide deck on your own, it contains more technical information for implementers
... it should be enough for people to come up with a PoC on their own

Ted: is there more of a requirement writeup instead of slides that can be shared? any other materials as well, if Caruso or Fraunhofer are willing to make code or access to an instance available that would be welcome

Dominik: the foundation of this is Oauth. there are really no more details other than in the second set of slides really necessary to come up with a PoC

Ulrich: our current PoC is not accessible from outside

Dominik: we could perhaps put the code in /github with instructions or make an instance available

Ted: my preference would be for code. that way people can also start combining things such as Benjamin's VSS SPARQL server

Tim: I can offer AWS host if that would help

Ted: that would be great, I can also get us VM at MIT at no cost. for the code feel free to use our github repo or your own. as long as things are referenceable on the web I can invite other people to look at them and provide their perspective. consent capture for example is not auto-specific

Ulrich: what would be helpful is coming up with end to end use case we want to address
... it could be predicative maintenance or any realistic story

Ted: I will include in minutes logistics for where people can put code and use cases if you want to use W3C infrastructure

Glenn: agree and will work on that with Geotab and Neutral Vehicle

[adjourned]