23G/ Discovery: Solving the Kobayashi Maru of SSI
Discovery: Solving the Kobayashi Maru of SSI
Thursday 23G
Convener: Daniel Hardman (daniel.hardman@evernym.com)
Notes-taker(s):
Tags for the session - technology discussed/ideas considered:
#discovery, #ssi, #privacy, #agents, #tor, #mediation, #data-ownership, #regulation
Discussion notes, key understandings, outstanding questions, observations, and, if appropriate to this discussion: action items, next steps:
How to allow discovery of personal information without running afoul of privacy, regulation, consent problems, and so forth is a conundrum in the context of SSI. We seem to have an unsolvable problem.
One way to address the tension is to change the rules. An important way to do this is to begin thinking of discovery as a collaborative effort that requires active participation of both parties (the party wanting to find, and the party wanting to be found). This is how Tinder works; both parties have to swipe right before any connection can be made. We could apply this same concept to discovery of DIDs or any other SSI data, such that a party wishing to be found is not looked up in an index, but rather is found by querying them directly. A decentralized matching service for discovery can be built that has cool properties that we haven't previously imagined.
Slides are here: https://j.mp/3obLky1
Paper with more technical details about the concepts: http://j.mp/ppred-paper
Chat transcript
From Neil Thomson : No good deed goes unpunished...
From Nathan_George : The Loan Officer Protocol needs to make it into the notes
From John Court : Isn’t Catch22 the original version of all these things ? “a paradoxical situation from which an individual cannot escape because of contradictory rules or limitations. The term was coined by Joseph Heller, who used it in his 1961 novel Catch-22.”
From RuffTimo : Love Star Trek. Brings back great memories. :)
From windley : This is a good idea for some of the chat spaces. We could just stream Star Trek in them. :)
From John Court : He cheated
From Jacob Siebach : There is a great clip from the game Star Fleet Academy where you can hack the simulator and the Klingons treat you well. Extremely funny.
From RuffTimo : Love IIW! Never know what you're gonna experience. :)
From Alan Karp : pre-Covid
From Ken Adler : If this was in Philipedlphia… I’m in the picture
From Jacob Siebach : "Please state the nature of the medical emergency."
From Nader Helmy : I think the example applies to any makeshift community
From Orie Steele : Resistance is futile
From RuffTimo : ^^ Perfect. :)
From Gabe Cohen : Or even worse…a PhD
From Bob Wyman : Being a doctor doesn’t ensure your willingness to help. In some states, being a “good samaritan” can present significant legal risk.
From Orie Steele : AshleyMadison.com ; BigData leads to information asymmetry
From Wayne Chang : ^ a good point I heard at IIW about ashleymadison is that the lawsuits didn’t fully capture the 2nd and 3rd order damages, such as suicides
From Orie Steele : TechnoViking… great story
From Vic Cooper : information asymmetry means that the market is ripe for disruption
From Orie Steele : See https://en.wikipedia.org/wiki/Techno_Viking ; Standing still is the best way to die in any FPS
From drummondreed : +1
From RuffTimo : Yeah, we quickly grew out of that perspective, thankfully. :) Maybe should've changed the name though...
From David Luchuk : Thank you for calling TechnoViking to mind. Truly.
From Orie Steele : Theory of negativity, is excellent branding ; Slack is another good example
From mitfik : Broadcasting discovery mechanism, you don't search but you let others to find you because of the need which you have.
From Ken Adler : Unless you pay Tinder Platinum…. Message before match :)
From Bob Wyman : Asking “Is there a Doctor in the house?” is polling. Polling is inefficient…
From Orie Steele : This is an example matching / clearing houses… there is entire division of economics dedicated to efficient matching of participant preferences
From Wayne Chang : In multi-sided markets you often have some legs that must be subsidized to make the model work.
From Ken Adler : yes
From Orie Steele : See also: https://en.wikipedia.org/wiki/Matching_theory_(economics)
From Nathan_George : (Right to be forgotten issues)
From Wayne Chang : Amazing book on market design https://www.amazon.com/Who-Gets-What-Why-Matchmaking/dp/0544705289
From Orie Steele : Everyone’s favorite database of int devices… https://www.shodan.io/ ; IoT *
From Wayne Chang : https://twitter.com/internetofshit
From Orie Steele : What you want is to build indexes of authorized service providers… which you can ask interactively.
From Ivan Temchenko : inter-hub gossips?
From Orie Steele : And be careful about what you make publically crawlable
From Bob Wyman : I think what you want is “Prospective Search” not the “Retrospective Search” that you are describing. i.e. let people publish the queries that they are willing to respond to. See: https://en.wikipedia.org/wiki/Prospective_search
From Orie Steele : Sharing an email or twitter handle is an invitation to be contacted.
From Ken Adler : Similar to buyer intent broadcasting
From Wayne Chang : I wish I could tell amazon when I move as opposed to it trying to guess that about my life
From windley : So this session should have been titled “Tinder for DIDs”
From Orie Steele : lul
From drummondreed : That’s a tweet there, Phil. Dare ya ;-)
From Dan Robertson (he/him) : Wow, Phil also a marketing savant... 😂
From RuffTimo : This could be a breakthrough for matching job candidates with jobs.
From drummondreed : Indeed
From RuffTimo : (without LinkedIn, Indeed, or other intermediary)
From Tyler @ Evernym : no Drummond, withOUT Indeed ;)
From Markus Sabadello : This example can even discover people who want a platypus as a pet AND as food!
From mitfik : basically this is what for we build "social platforms" those platforms are matchers in that case
From Wayne Chang : @markus RDF is truly powerful
From RuffTimo : @Markus Platypi are delicious. ;)
From Bob Wyman : This is “cross matching,” a combination of “prospective and retrospective” search.
From windley : https://twitter.com/windley/status/1319376968022855680
From Orie Steele : The problem is that greedy bots want to match with everything, and sybil lets them map the entire space…. These kinds of systems get terribly complex… they are almost always solved by relying on a trusted centralized clearing house.
From Judith Fleenor : There must be a way to revoke your desire to be matched by each key.
From Wayne Chang : @judith, would that requirement be dependent on how much info is exposed via match request?
From RuffTimo : Lovin' this... very exciting stuff, addressing a VERY tricky and problem.
From John Court : There was a DEC research project in the 90s called Each-to-Each which this reminds me of, except it was centralised matching and not diffused across Many to maintain privacy.
From Judith Fleenor : @wayne exactly
From drummondreed : He took the dare! Just retweeted, Phil.
From Gabe Cohen : +1 Orie this is incredibly noisy
From mitfik : Spamers would love it :)
From Orie Steele : Its a super hard problem in economics
From Wayne Chang : You could add friction to match requests ; The microeconomic problem you’re talking about is market congestion
From Adrian Gropper : Starting to sound like the Apple Google COVID proximity scheme
From drummondreed : Reputation plays a major role here
From Wayne Chang : https://hbr.org/2007/10/the-art-of-designing-markets
From Orie Steele : Yes, you need to make queries costly ; And joining expensive
From Gabe Cohen : I will pay 3 btc to find my dog the perfect pal
From Orie Steele : lul
From Wayne Chang : Are we queueing
From Adrian Gropper : q+
From drummondreed : Search bounties that are payable on success could work nicely
From Bart Suichies : so how is this different than MPC?
From Tyler @ Evernym : I require payment to be matched for certain things.
From Judith Fleenor to Daniel Hardman (Privately) : Love how organized this presentation is WELL Done presentation… and loved the fun elements such as the title and opening videos.
From Orie Steele : https://ieeexplore.ieee.org/document/6765218
From McCown : Doesn’t “Kobayashi Maru” imply success by surreptitiously reprogramming the system? ;-)
From Orie Steele : ^ there is a lot of academic research in this area
From Tyler @ Evernym : i.e., pay me first in order to have the privilege of matching with me
From Vic Cooper : I love this in the context of a telecommunications platform. Might solve the problem of who can ask to connect to me
From Wayne Chang : @vic I think this was earn.com’s business model prior to their acquisition & pivot
From Judith Fleenor : @VIC wouldn’t it be nice to solve the Robo Call issues…
From Wayne Chang : Pay to talk to someone, and the someone could direct that payment to a charity for a less socially awkward outcome ; But it solves the problem of lowering the noise floor
From Vic Cooper : yes or have some sort of attention token so that there is a cost to connect to me and I can decide on the price
From Adrian Gropper : context, crypto, agency
From Wayne Chang : That’s just setting your own price but with more steps
From David Huseby : key agreement “prekeys” from Noise can require the initiating party to do some computation work to calculate their half of the 3DH
From Wayne Chang : hashcash++
From David Huseby : by using something like PBKDF
From Gabe Cohen : Tl;dr advertising with didcomm
From Wayne Chang : I like the idea of sending a private key of a cryptocurrency account with a small amount
From David Huseby : so a person who wants to be discovered publishes their half of the prekey with the hashcash like challenge that has to be solved to calculate the other side
From Wayne Chang : See also some cool discussion: https://github.com/decentralized-identity/didcomm-messaging/issues/66
From Nathan_George : Sounds like some interesting papers could be linked in the notes?
From David Huseby : nature has interesting mechanisms for discovery
From Orie Steele : Slime mold is my favorite organic search system
From Wayne Chang : Vote rename TCP/IP to ANT PROTOCOL
From Bart Suichies : @David: pheromones?
From Orie Steele : There are also randomized solutions to traveling salesman that are based on ants.
From Wayne Chang : I love all sentences ending in “that are based on ants"
From David Huseby : @Bart, yes, scent marking, compressing space like salmon spawning in rivers ; those are different ways ; preventing enumeration is key. ; that’s pretty much all this idea prevents
From Bart Suichies : https://biomimicry.org/solution/slant/
From Adrian Gropper : Hence the need for powerful agents; Every query is a request with three components: ; - Claims ; - Scope ; - Purpose
From Daniel Hardman : @Adrian: +1
From Bart Suichies : powerful and logically distributed
From drummondreed : I’m a huge fan of agent-based discovery. That’s actually how Kirk solved the “is there a doctor in the house?” problem. An agent answered (in that case, the agent was a human)
From RuffTimo : Very cool point, Daniel, about matching with both privacy and verifiability of attestations.
From Orie Steele : If the matches collude they can recover related keys, which almost gives them the same information in a VDR
From Adrian Gropper : +1 Orie
From Nader Helmy : This solves a really important problem, gives sovereign individuals the ability to organize and cooperate in a way that’s typically reserved for formal organizations with their own governance
From Bob Wyman : Actually, matching is often more efficient if it is NOT sharded. But, it is easier to scale if shared.
From drummondreed : Nader, +1
From Bob Wyman : But, it is easier to scale if sharded. (autocorrect error)
From RuffTimo : Super insightful session, Daniel, and important for the space. IMO this will spawn a lot of great discussion in this community and elsewhere. Kudos!
From windley : Back in the day (2012), Drummond’s company (Respect Network, Cntl-Shift, and my company (Kynetx) built a pico-based, VRM-like system for Innotribe (SWIFT) that did this kind of matching. Here’s a video showing it (that Heather Vescent helped with). https://vimeo.com/51827693
From Nader Helmy : Social cooperation is at the core of what makes us human, its a failure if we dont solve this problem in a mainstream and easily adoptable way
From drummondreed : Wow, Phil, you found the video! Cool!
From windley : Blog search
From Nathan_George : Reminder that Wayne and Adrian are on the queue
From drummondreed : Intentcasting! That’s what I’ve been trying to remember this whole session.
From Adrian Gropper : q+ to point out this is important to inform our community'swallet vs. agent issues
From RuffTimo : Ooh... glad you brought up intentcasting, Phil... Isn't this fundamentally intentcasting?
From drummondreed : What Daniel is proposing is a very sophisticated version of intentcasting (and intent querying)
From RuffTimo : Interesting to think of how this could be used for nefarious purposes as well, allowing bad guys to find/connect... but as with all new tech, it can be used for both good and bad.
From drummondreed : This is very true. And believe me, some governments will remind us about that. Loudly.
From Nathan_George : “A horse, a horse! My kingdom for a horse!”
From David Huseby : IPv6 gives us an interesting opportunity. If we all have our own /64 then we have 64 bits to use for mapping content addresses to IP endpoints in our subnet. ; with mobile ipv6 it will be possible to literally broadcast ping all devices on a given access point to an endpoint in all devices
From David Huseby : so if I hash dog, then concatenate sushi and hash that to get the endpoint in my subnet, I can have my device listen there and then ping all other devices on that endpount
From Vic Cooper : http://bit.ly/ppred-paper
From Wayne Chang : @Dave I wonder what are some ways to make it more expensive to enumerate active possible matches by brute forcing those concat-hash queries
From David Huseby : it uses IP addresses to “compress” content addresses to an IPv6 address “bottom half” that I can use to ping devices with
From windley : More on intent casting: https://www.windley.com/archives/2012/06/buying_a_motorcycle_a_vrm_scenario_using_personal_clouds.shtml
From David Huseby : @Wayne, default end point hands out prekeys with hashcash problems ; my device then drops all packets that aren’t the correct half of a handshake
From Wayne Chang : Ooh nice, I was suspecting some hashcash component but didn’t think of the prekey by the gateway ; cool
From David Huseby : so devices can essentially name their price
From Wayne Chang : Yeah I also like that the gateway isn’t doing the filtering ; The intermediary here is an “anonymizing” endpoint though
From David Huseby : @Wayne the noise 3DH zero round trip handshakes requires the connecting party to get the receiving party’s prekey and then they calculate the other half of the key agreement and encrypt the first packet
From Wayne Chang : To disintermediate this would need something like onion routing
From David Huseby : but a hashcash problem can also be mixed in so that the connecting party has to expend non-trivial computation cost
From Nathan_George : Note: Bob then Robert then Trev are still on the queue
From RuffTimo : Idea: those here most interested in this connect with Daniel for purposes of continuing the discussion after IIW, and push this forward without delay. This is a big piece of the SSI puzzle, IMO.
From Bob Wyman : How do we “connect with Daniel?”
From Laura J : Where can we find Dave’s papers?
From Wayne Chang : @bob hahahaha
From drummondreed : daniel.hardman@evernym.com
From Wayne Chang : But Drummond, where do I swipe right
From Daniel Hardman : “Connect with Daniel”: daniel.hardman@evernym.com
From RuffTimo : There's this discovery protocol for connecting with Daniel...
From Laura J : haha
From Bruce Conrad : Solve the problem and Daniel will come to you
From John Court : What is the load for listening on many addresses that way ? Energy and chip limits ?
From Orie Steele : Proof of work or proof of payment
From Colin Jaccino : Seems like a multicast scenario might work for this example as well
From Trent Larson : @wayne @bruce :-D Pure gold.
From Nathan_George : Queue: Robert then Trev ; Queue: Robert then Trev then Trent
From Nader Helmy : Hey Dave where can we find your papers?
From Orie Steele : Yes, one way of handling information compression is to register algorithms instead of specific tags ; And then deterministically generate the tags from the algorithm ; Btw this is how CBOR-LD works
From David Huseby : https://link.medium.com/iGXlkYcINab
From Joe Andrieu : Thanks, Daniel!
From David Huseby : principles of user sovereignty ; https://link.medium.com/TDdUMYeINab ; https://link.medium.com/IGPEVjgINab ; those three papers are ; principles of self sovereignty ; a unified theory of decentralization ; the web was never decentralized
From Nader Helmy : +1 ; UX is a huge part of this working
From David Huseby : another idea would be a “scavenger hunt” where searchers have to do proof of work to find the location of a piece of data that then gives them another problem to find the next ; to index all data they would have to tons of work
From Bob Wyman : Technology can’t overcome prejudice.
From Nathan_George : Private connections means less consequences for refusing to cooperate — interesting
From Wayne Chang : @Dave capture the flag to use the service
From David Huseby : @Wayne...yes!
From Bob Wyman : A bigot with a computer is still a bigot.
From Phil Wolff : SEVEN MINUTE WARNING
From Nader Helmy : @Bob there’s bias in all technology, incentives really matter at scale
