Re: [Din] WSJ article on Identity and Blockchains

[Jon Crowcroft <[e-mail]>]

this al sounds excellent - socially rooted, community operated communities
- all good - we could think of building instances of this on the
infrastructure we've been constructing (i know there are lots of other
people building similar, so this is a bit of gratuitous
self-aggrandizing:), viz
http://www.databoxproject.uk/

there are some decentralisation movements in the certificate space (CA
transparency) as wel as self-certifying address allocation. so I assume
there's some hope for using these to bootstrap the platforms-on-demand,
safely for  communities you're envisaging -

we're definitely on same page here!


On Sun, Apr 8, 2018 at 3:06 PM, Thomas Hardjono <[e-mail]> wrote:

>
> Thanks Jon.
>
> Aside from the obvious goal of cutting through the current hype about
> "Self-Sovereign Identity" and how Blockchains (BC) will achieve "my
> sovereignty",  we wanted to communicate to the general  reader that the
> problem of "identity" is not so simple.
>
> >>> so you _are_ your social network...in terms of trustworthy
> identity...sure...
>
> No, not so simple, particularly when the social network is FB :-)
>
>
> >>> there's two problems with this though in details...
> >>> i.e. how we build on this idea technically in the Din context...
>
> >>> 1/ its still dependent on technologies,
> >>> and there's a seperate issue of why we trust them to proxy our social
> net -
> >>> i certainly would it find it hard to trust any social media app,
> running on a cloud platform,
> >>> using a smart mobile device, to vouch for all these friends &
> colleagues - too many layers, too many
> >>> vested interests, too much Cambridge Analytica :)
>
> You've hit the nail, as usual - thank you Jon :-)
>
> The question is how to capture/use/protect the human-to-human interaction
> as the basis for trust ("local trust").
>
> Here is an example. So my local town here in the US has a private
> mailing-list for local residents only. I have to attend town meetings
> in-person before I can get on to that mailing-list.
>
> There is no reason why my local town cannot run their own "social
> platform" (we are in the age of VMs, Containers, Clouds).  It must legally
> owned by the town/community, and available to residents only (i.e. who pay
> local tax), etc.
>
> How do we create next-gen architectures/protocols/legal that support local
> communities to interact  on their own "social platform" , and become the
> starting point for "trust" that can scale out (called "local trust" in the
> WSJ paper).
>
>
> I think there is huge role for DIN to help. For the future distributed
> Internet infrastructure, is it possible to have the following:
>
> -- Can my town obtain a "social-platform-on-demand" from a cloud service
> provider.
> -- Can I legally prevent the service-operator from "making use" of my data
> (yes they still may be able to steal my data).
> -- Can I retain legal ownership of my data (same with my town).
> -- Can I encrypt my data on my town's social-platform so that the
> service-operator doesn't steal it.
> -- How to do key management.
> -- How can I use legal smart-contracts to bind the service-operator.
> -- Etc. etc. lots more questions for DIN RG.
>
>
>
> >>> 2/ I though many people in the security community were moving away from
> >>> proving identity, towards systems that prove entitlement (i.e.
> credentials
> >>> are on a need-to-know basis, so if you were say 19, you don't need to
> say yur age or show id,
> >>> but you can't buy a drink in cambridge MA, but you can in cambridge,
> UK :)
>
> Yes, correct. Currently people are focusing on proving entitlements via
> signed assertions. Great, no issue there. But the real question is about
> the *data* and *algorithms* behind those assertions (that impact my life,
> like credit-score):
>
> -- How do I know that an organization has subject-data about me (usual
> questions of "how, what, when, what are you doing to my data").
> -- What is the provenance of the data in their possession  (Did they get
> it from a data-aggregator? Do they even know?)
> -- Does the data set have in-built bias? (not intentionally)
> -- Is the data "fair"  (fairness in the sense of machine learning).
> -- What is the accuracy of the algorithms computed on my data. Does it
> propagate the bias in the data.
> -- Do I get to see the algorithms that impact my life.
> -- Can I hire an expert to provide objective review of these algorithms.
> -- etc. etc. etc.
>
>
>
> >>> bootstrapping something from a BC to provide the credentials is also
> problematic, in that
> >>> BC needs a PKI to know whether nodes are not sybils, spoofs, etc, so
> we have a circular dependance, no?
>
> Yes, agree.  There is need to revisit PKI again, maybe opportunity in DIN
> RG to improve it.
>
> Hashing an identifier (and back-links) to a BC doesn't help the
> identity/trust problems.
> Hashing a pub-key to a BC also doesn't add much.
>
>
> >>> maybe i missed an important step, if so, sorry!
>
> Nope, you are on the right track :-)  Hope you can help solve some these
> problems.
>
> Best.
>
>
> -- thomas --
>
> ________________________________________
> From: Jon Crowcroft [[e-mail]]
> Sent: Sunday, April 08, 2018 6:38 AM
> To: Thomas Hardjono
> Cc: [e-mail]; Jon Crowcroft
> Subject: Re: [Din] WSJ article on Identity and Blockchains
>
> very nice article...
>
> so you _are_ your social network...in terms of trustworthy
> identity...sure...
>
> there's two problems with this though in details...
> i.e. how we build on this idea technically in the Din context...
>
> 1/ its still dependent on technologies,
> and there's a seperate issue of why we trust them to proxy our social net -
> i certainly would it find it hard to trust any social media app, running
> on a cloud platform,
> using a smart mobile device, to vouch for all these friends & colleagues -
> too many layers, too many
> vested interests, too much Cambridge Analytica :)
>
> 2/ I though many people in the security community were moving away from
> proving identity, towards systems that prove entitlement (i.e. credentials
> are on a need-to-know basis, so if you were say 19, you don't need to say
> yur age or show id,
> but you can't buy a drink in cambridge MA, but you can in cambridge, UK :)
>
> bootstrapping something from a BC to provide the credentials is also
> problematic, in that
> BC needs a PKI to know whether nodes are not sybils, spoofs, etc, so we
> have a circular dependance, no?
>
> maybe i missed an important step, if so, sorry!
>
>
> > Folks,
> >
> > I thought to share this WSJ article with the DIN group. Relevant in the
> > light of recent interest in using BC for identity.
> >
> > Advance apologies if it offends some people :-)
> >
> > https://blogs.wsj.com/cio/2018/04/03/digital-identity-
> is-broken-heres-a-way-to-fix-it/
> >
> >
> > Below is a link to a PDF version.
> >
> > http://hardjono.mit.edu/sites/default/files/documents/WSJ_
> Digital_Identity_is_Broken.pdf
> >
> >
> > Best
> >
> > -- thomas --
> >
> > _______________________________________________
> > Din mailing list
> > [e-mail]
> > https://www.irtf.org/mailman/listinfo/din
> >
>