7 ms·

Brave developer here. Great question! Brave Ads are entirely opt-in, and work via on-device machine-learning. The machine-learning bits study your browsing habi

Brave developer here. Great question! Brave Ads are entirely opt-in, and work via on-device machine-learning. The machine-learning bits study your browsing habits in a private, non-leaky manner. Once a day, your device downloads an aggregate catalog of many ad options to be studied locally. If/when an ad is found that fits your interests (as inferred by your browsing habits), the ad is displayed as an OS notification and 70% of the ad revenue is deposited into your in-situ wallet. So on the topic of targeting, Brave is able to deliver a better experience for Advertisers and Users, without the need to leak user information across the Web. The on-device ML bits learn about their user over time, delivering a better experience with maturity. I hope that helps!

seren 7y ago
Thanks for chiming in. So from the FAQ : https://brave.com/faq/#all-ads-blocked https://brave.com/faq/#all-ads-blocked > Each ad request is anonymous, and exposes only a small subset of the user’s preferences and intent signals to prevent “fingerprinting” the user by a possibly unique set of tags. It means that advertisers still get some feedback on the "intent signals" of the users that have seen their ads ? So it is private in the sense you cannot be uniquely identified but some of your intents are somewhat leaky ?
- jonathansampson 7y ago
  Not quite, I think that FAQ item needs to be updated. The model works like this: The user opts-in to Brave Rewards/Ads. This kicks off a machine-learning model that begins to study the user's habits and interests. On a daily basis, the user downloads a regional ad catalog (as to other Brave users in their region). This catalog contains numerous ad options, which your device studies for relevance. If/when your device identifies an ad within the catalog that might be of interest to you, it displays the ad as an OS notification. At this time, 70% of the ad revenue is deposited into your in-situ wallet. At this point, the user has made absolutely no contact with an advertiser. If the user chooses to click on the ad, it is opened in its own tab in the Brave Browser. This tab, like all others, is subject to Brave's default security/privacy settings. No third party trackers, etc. The advertiser will know that you are interested in their product/service, because you clicked the ad. But no other information about you is given to them; only what can be inferred from standard first-party browsing online. Brave aims to keep your data private, and on-device. It is never leaked to us, or anybody else. You are sovereign over your data, and you decide with whom it is shared. That's our goal, always.
  - seren 7y ago
    Thanks that's very clear and helpful.
  - tomcam 7y ago
    Man, just copy and paste that answer into the FAQ as is. Nailed it.
  - CrendKing 7y ago
    > At this time, 70% of the ad revenue is deposited into your in-situ wallet. Brave needs to know when and which ad has been shown to the user, to deposit that ad's revenue right? That means whenever an ad is shown "locally", it still needs to contact Brave server with user's identity behind of some encryption. Brave then shows the proof of ads being presented to users to ad provider to get revenue. Sure, the user's interests are studied locally (and inefficiently because it runs on user's computer), but the result feeds back to Brave. How does that improve privacy? At some point, Brave may either leak the data accidentally or decide to sell the data, because the data, which capture each uniquely identifiable user's interests, is valuable. Or am I misunderstanding something here?
    - sexisfun 7y ago
      How do you purposefully leak data and get paid for it without a paper trail?
    - BrendanEich 7y ago
      We use a variant of privacy pass (Chaum blind signatures): https://github.com/brave-intl/challenge-bypass-ristretto https://github.com/brave-intl/challenge-bypass-ristretto The key is no user identifiable events and no ability to link events for a given user together. Advertisers want authentic aggregate results — they don’t want (at first or in bulk, also not legal in many places) user ids. We built an authentic but anonymous ad system.
jayd16 7y ago
So it has to download every possible ad (to mask my identity/preference)?
- jonathansampson 7y ago
  Catalogs are regional, so you aren't downloading every possible ad. You're downloading a ~3MB or so file that has ads for your region. Others in your region download the same catalog, which prevents fingerprinting.
im3w1l 7y ago
How do you prevent ad fraud while respecting privacy?
- jonathansampson 7y ago
  We don't rely on extension and/or DOM APIs like other approaches—we're the browser. It's much harder to fool the browser. Additionally, because our models are built on client-side operations, analysis on the client can be done without the need to leak any data from the user's session. That's about as much as I'm happy to discuss. Fraud prevention is one of the few opaque areas in our project, for good reason. I hope that helps!
  - cameronbrown 7y ago
    So fraud prevention is closed source? How is that fair?
asdkhadsj 7y ago
Does the viewing of an ad leak data? Eg, if they know my IP viewed an ad, and they were targeting males 15-25, wouldn't that leak the data? Are there systems in place to prevent this sort of data leak as well?
Circuits 7y ago
Totally confused... why should it matter to me, the user, if it's you tracking me or if it's Google tracking me?
vesche 7y ago
I would never opt-in to ads even if there was some upside. That's like asking if I'd like to see billboards on a beautiful drive through Colorado and in return you'll throw some change in my center console. I'll just take the beautiful drive and you can keep the cash. > The machine-learning bits study your browsing habits in a private, non-leaky manner. I think everyone's been around the internet long enough to know that nothing is private forever and everything has leaks. Relevant: https://twitter.com/briankrebs/status/1045091640480804864 https://twitter.com/briankrebs/status/1045091640480804864
- kerkeslager 7y ago
  > I would never opt-in to ads even if there was some upside. That's like asking if I'd like to see billboards on a beautiful drive through Colorado and in return you'll throw some change in my center console. I'll just take the beautiful drive and you can keep the cash. This, and the fact that Brave has decided to get in bed with advertisers at all seriously undermines their credibility.
  - BrendanEich 7y ago
    Only if you assume we enable tracking for Brave Ads. But we do not. Please look deeper, otherwise you're just writing false stuff based on assumptions.
    - kerkeslager 7y ago
      No, nothing in my post is false. I knew how your ad model works when I wrote that post. You're in bed with advertisers because you're taking money from them and providing them a service. Look, I don't have anything against you personally. I'm sure you're a good, kind person with pure intentions. But tracking or not, advertising is not the way to fund a browser that serves users. You'll always be beholden to advertisers for your cash flow, and you're always going to be making choices between prioritizing users or prioritizing your income. Sometimes the choices will seem minor and it won't be clear which direction to go, and you'll compromise and the focus on users will be eroded. And at some point the choice might be between screwing over users and shutting the doors of your company due to lack of funding or something else. Historically, some people have chosen users[1] but most have chosen to stay open. I hope you prove me wrong, I really do. But you'll excuse my skepticism at this point. [1] https://en.wikipedia.org/wiki/Lavabit https://en.wikipedia.org/wiki/Lavabit
sharemywin 7y ago
When do you build a digital assistant/agent that works on a similar model. I would even volunteer some data for off machine I could choose which data goes. Aka speech data for training an agent.

seren 7y ago

Thanks for chiming in. So from the FAQ : https://brave.com/faq/#all-ads-blocked https://brave.com/faq/#all-ads-blocked > Each ad request is anonymous, and exposes only a small subset of the user’s preferences and intent signals to prevent “fingerprinting” the user by a possibly unique set of tags. It means that advertisers still get some feedback on the "intent signals" of the users that have seen their ads ? So it is private in the sense you cannot be uniquely identified but some of your intents are somewhat leaky ?

jonathansampson 7y ago
Not quite, I think that FAQ item needs to be updated. The model works like this: The user opts-in to Brave Rewards/Ads. This kicks off a machine-learning model that begins to study the user's habits and interests. On a daily basis, the user downloads a regional ad catalog (as to other Brave users in their region). This catalog contains numerous ad options, which your device studies for relevance. If/when your device identifies an ad within the catalog that might be of interest to you, it displays the ad as an OS notification. At this time, 70% of the ad revenue is deposited into your in-situ wallet. At this point, the user has made absolutely no contact with an advertiser. If the user chooses to click on the ad, it is opened in its own tab in the Brave Browser. This tab, like all others, is subject to Brave's default security/privacy settings. No third party trackers, etc. The advertiser will know that you are interested in their product/service, because you clicked the ad. But no other information about you is given to them; only what can be inferred from standard first-party browsing online. Brave aims to keep your data private, and on-device. It is never leaked to us, or anybody else. You are sovereign over your data, and you decide with whom it is shared. That's our goal, always.
- seren 7y ago
  Thanks that's very clear and helpful.
- tomcam 7y ago
  Man, just copy and paste that answer into the FAQ as is. Nailed it.
- CrendKing 7y ago
  > At this time, 70% of the ad revenue is deposited into your in-situ wallet. Brave needs to know when and which ad has been shown to the user, to deposit that ad's revenue right? That means whenever an ad is shown "locally", it still needs to contact Brave server with user's identity behind of some encryption. Brave then shows the proof of ads being presented to users to ad provider to get revenue. Sure, the user's interests are studied locally (and inefficiently because it runs on user's computer), but the result feeds back to Brave. How does that improve privacy? At some point, Brave may either leak the data accidentally or decide to sell the data, because the data, which capture each uniquely identifiable user's interests, is valuable. Or am I misunderstanding something here?
  - sexisfun 7y ago
    How do you purposefully leak data and get paid for it without a paper trail?
  - BrendanEich 7y ago
    We use a variant of privacy pass (Chaum blind signatures): https://github.com/brave-intl/challenge-bypass-ristretto https://github.com/brave-intl/challenge-bypass-ristretto The key is no user identifiable events and no ability to link events for a given user together. Advertisers want authentic aggregate results — they don’t want (at first or in bulk, also not legal in many places) user ids. We built an authentic but anonymous ad system.

jayd16 7y ago

So it has to download every possible ad (to mask my identity/preference)?

jonathansampson 7y ago
Catalogs are regional, so you aren't downloading every possible ad. You're downloading a ~3MB or so file that has ads for your region. Others in your region download the same catalog, which prevents fingerprinting.

im3w1l 7y ago

How do you prevent ad fraud while respecting privacy?

jonathansampson 7y ago
We don't rely on extension and/or DOM APIs like other approaches—we're the browser. It's much harder to fool the browser. Additionally, because our models are built on client-side operations, analysis on the client can be done without the need to leak any data from the user's session. That's about as much as I'm happy to discuss. Fraud prevention is one of the few opaque areas in our project, for good reason. I hope that helps!
- cameronbrown 7y ago
  So fraud prevention is closed source? How is that fair?

asdkhadsj 7y ago

Does the viewing of an ad leak data? Eg, if they know my IP viewed an ad, and they were targeting males 15-25, wouldn't that leak the data? Are there systems in place to prevent this sort of data leak as well?

Circuits 7y ago

Totally confused... why should it matter to me, the user, if it's you tracking me or if it's Google tracking me?

vesche 7y ago

I would never opt-in to ads even if there was some upside. That's like asking if I'd like to see billboards on a beautiful drive through Colorado and in return you'll throw some change in my center console. I'll just take the beautiful drive and you can keep the cash. > The machine-learning bits study your browsing habits in a private, non-leaky manner. I think everyone's been around the internet long enough to know that nothing is private forever and everything has leaks. Relevant: https://twitter.com/briankrebs/status/1045091640480804864 https://twitter.com/briankrebs/status/1045091640480804864

kerkeslager 7y ago
> I would never opt-in to ads even if there was some upside. That's like asking if I'd like to see billboards on a beautiful drive through Colorado and in return you'll throw some change in my center console. I'll just take the beautiful drive and you can keep the cash. This, and the fact that Brave has decided to get in bed with advertisers at all seriously undermines their credibility.
- BrendanEich 7y ago
  Only if you assume we enable tracking for Brave Ads. But we do not. Please look deeper, otherwise you're just writing false stuff based on assumptions.
  - kerkeslager 7y ago
    No, nothing in my post is false. I knew how your ad model works when I wrote that post. You're in bed with advertisers because you're taking money from them and providing them a service. Look, I don't have anything against you personally. I'm sure you're a good, kind person with pure intentions. But tracking or not, advertising is not the way to fund a browser that serves users. You'll always be beholden to advertisers for your cash flow, and you're always going to be making choices between prioritizing users or prioritizing your income. Sometimes the choices will seem minor and it won't be clear which direction to go, and you'll compromise and the focus on users will be eroded. And at some point the choice might be between screwing over users and shutting the doors of your company due to lack of funding or something else. Historically, some people have chosen users[1] but most have chosen to stay open. I hope you prove me wrong, I really do. But you'll excuse my skepticism at this point. [1] https://en.wikipedia.org/wiki/Lavabit https://en.wikipedia.org/wiki/Lavabit

sharemywin 7y ago

When do you build a digital assistant/agent that works on a similar model. I would even volunteer some data for off machine I could choose which data goes. Aka speech data for training an agent.