Metadata Leakage

It's a well-known fact that Nostr's NIP 04 DMs leak metadata. This seems like an obvious flaw, and has been pointed out as such many times. After all, if anyone can see who you're messaging and how frequently, what time of the day, how large your messages are, who else is mentioned, and correlate multiple separate conversations with one another, how private are your communications really?

A common retort repeated among those who "get" Nostr (myself included) is "it's not a bug, it's a feature". This hearkens back to the early days of the internet, when internet security was less than an afterthought, and social platforms throve on various permutations of the anonymous confessions-type app. How interesting to be able to flex to your friends about whom you're DMing and how often! Most conversations don't really need to be private anyway, so we might as well gamify them. Nostr is nothing if not fun.

In all seriousness though, metadata leakage is a problem. In one sense, Nostr's DMs are a huge improvement over legacy direct messages (the platform can no longer rat you out to the FBI), but they are also a massive step backward (literally anyone can rat you out to the FBI). I'm completely confident we'll be able to solve this issue for DMs, but solving it for other data types within Nostr might pose a bigger problem.

Social Content

A use case for Nostr I've had on my mind these last few months is web-of-trust reviews and recommendations. The same sybil attack that allows bots to threaten social networks has also been used as a marketing tool for unscrupulous sellers. NPS surveys, purchased reviews, and platform complicity have destroyed the credibility of product reviews online, just like keyword-stuffed content has ruined Google's search results.

Proof-of-work would do nothing to defend against this attack, because the problem is not volume, it's false credibility. The correct tool to employ against false credibility is web-of-trust — verifiable trustworthiness relative to the end user's own social graph.

This is a huge opportunity for Nostr, and one I'm very excited about. Imagine you want to know whether the vibro-recombinant-shake-faker (VRSF) will result in visible abs in under 6 days. Well, it has over 4 thousand 5-star reviews on Amazon, and all the 1-star reviews are riddled with typos and non sequiturs. So it must work, and make you smarter into the deal! Well, sadly no, visible abs are actually a lie sold to you by "big gym".

Now imagine you could find your three friends who fell for this gyp and ask them what they thought — you might just end up with a lower average rating, and you'd certainly have a higher level of certainty that the VRSF is not worth the vibra-foam it's molded from.

This same query could be performed for any product, service, or cultural experience. And you wouldn't be limited to asking for opinions from your entire social graph, it would be easy to curate a list of epicureans to help you choose a restaurant, or trusted bookworms to help you decide what to read next.

Currently, big tech is unable to pull this off, because Facebook won't share its social graph with Google, and Google won't share its business data with Facebook. But if an open database of people and businesses exists on Nostr, anyone can re-combine these silos in new and interesting ways.

Notes and other Spies

So that's the pitch, but let's consider the downsides.

An open social graph coupled with recommendations means that not only can you ask what your friends think about a given product, you can ask:

• What a given person's friends think about a product
• What kind of person likes a given product
• How products and people cluster

That last one in particular is interesting, since it means you could find reasonable answers to some interesting questions:

• Does a given region have fertility problems?
• What are the political leanings of a given group?
• How effective was a particular advertisement with a given group?

This is the kind of social experiment that has historically earned Facebook so much heat. Democratizing this data does not prevent its correlation from being a violation of personal privacy, especially since it will be computationally expensive to do sophisticated analysis on it — and the results of that analysis can be kept private. And to be clear, this is a problem well beyond the combination of social information and public reviews. This is just one example of many similar things that could go wrong with an open database of user behavior.

Not to put too fine a point on it, we are at risk of handing the surveillance panopticon over to our would-be overlords on a silver platter. Just as walled gardens have managed us in the past to sway political opinion or pump the bags of Big X, an open, interoperable content graph will make building a repressive administrative state almost too easy.

Let's not give up just yet

So what can we do about it? I want a ratings system based on my social graph, but not at the expense of our collective privacy. We need to keep this threat in mind as we build out Nostr to address novel use cases. Zero-knowledge proofs might be relevant here, or we might be able to get by with a simple re-configuration of data custody.

In the future users might publish to a small number of relays they trust not to relay their data, similar to @fiatjaf's NIP-29 chat proposal. These relays might then support a more sophisticated query interface so that they can answer questions without revealing too much information. One interesting thing about this approach is that it might push relays towards the PWN model BlueSky uses.

Not all data needs to be treated the same way either, which would give us flexibility when implementing these heuristics. Just as a note might be either broadcast or sent to a single person or group, certain reviews or other activity might only be revealed to people who authenticate themselves in some way.

Like so many other questions with Nostr, this requires our concentrated attention. If all we're doing is building a convenient system for Klaus Schwab to make sure we ate our breakfast bugs, what are we even doing?