Assuming you are running an ad network you kind of have to in order to prevent ad fraud. Also by reducing that data you know about someone's interests is the knowledge that they have visited a site at least you will not be able to pick as good of an ad compared to if you had more data.
In that scenario, it seems like advertisers would pick up on that pretty quickly when they realize the conversion rate on that supposed traffic is terrible and doesn't warrant the inflated price. In the case they're using an ad network, the network could ban the page owner from their network if they see this pattern from them. Since page owners are materially benefiting from the network, proof of identity should be (probably is? I don't work in the space) applied between the page owner and network to prevent repeat fraud via identity laundering.
> you will not be able to pick as good of an ad compared to if you had more data.
In theory I lean towards agreeing. I was arguing for tech-powered hyper-personalized ads back when I was studying advertising 2008-2012 (and did a bit of "stealth marketing" in that period where I built relationships with bloggers to share our product before the term "influencer" had hit the mainstream vocabulary).
In practice, advertisers do not personalize ads. Facebook has become pretty good about selecting ads that map to my interests thanks to the reach of their spy network, but the ads themselves still aren't personalized at all (they take my interests into account, but not my spending history to realize that I don't have the budget for what they're trying to sell me) and my conversion to a sale because of them is still very, very low.
Instead of selling space by impressions or clicks, we use length of time (monthly) and find it to be a good way to prevent ourselves from trying to game impressions with clickbait or clicks with fake users.
1. Page owner / Ad network / Ad space auction market middleman fakes clicks to get click revenue
2. Page owner's rival fakes clicks to devalue ad spots
3. Advertiser's agency fakes clicks to make numbers go up
4. Advertiser's rivals fake clicks, to waste advertiser's budget
5. Ad networks 'accidentally' classifying legitimate clicks as fraud, to reduce payouts to page owners.
1. Clicks are not charged for, so increasing clicks does not change the cost.
2. The only question available is how much you pay and how much you make. Page owner's rival could attempt to confuse you but there's not actually a difference between paying $1000 for 0.1% of ads shown in July, getting 100 clicks out of 10000 impressions and making $2000 vs paying $1000 for 0.1% of ads shown in July, getting 10000 clicks out of 19900 impressions and making $2000.
3. Shorter 2: Numbers are irrelevant unless they have a dollars/euro/yen/pound sign before them.
4. Clicks to not come out of the advertisers budget.
5. Surveillance advertising enables this fraud by attempting to distinguish between real and fake traffic. If you are paying for something that is robust to that, then the fraud is irrelevant. In particular, it doesn't matter if 0.1% of ads displayed to bots are yours and 0.1% of ads displayed to humans are yours if you've paid for 0.1% of ads to be displayed to be yours using a random rotation.
The only thing like any of these hacks that seems relevant is that a very crafty adversary might be able to determine your traffic rate and identify the timeframe they need to request to particularly target a certain ad, making more of the 0.1% of displays get shown to a bot. But this is easily handled by using a secure random rotation rather than say a round robin rotation. In any case, I can independently verify the efficiency of my spend by getting a sim card that the ad network doesn't know about and repeatedly reloading the page for a month. If I receive only 0.01% of displays because some third party or the network is trying to defraud me I now have pure evidence that they aren't upholding their side of the deal. If they try to verify humanity before deciding what to do, then they can just say "no, we identified these requests as coming from a bot and graciously didn't display your ad to them". What can you do? The existence of surveillance is not in the interest of the advertiser who is concerned about fraud.
(If you don't trust your network, any fraud on their part isn't going to get solved of course. But then having less data is an advantage. There will be fewer variables so it will be easier to catch, and the capital requirements will be simpler --- code and business partners vs code, business partners and user surveillance data --- so it will be easier for a trustworthy competitor to emerge.)
Harder to pull off than advertising at people who might actually want to see the ads? More risk and complication than the growing backlash against advertising in general entirely because of shady advertising practices? More risk and complication than having to keep track of various countries' and states' laws re; privacy?
> "For example, what if a web master decides they don't want to have ads on their site anymore. Whoever just paid for that space gets screwed."
Existing contract law already covers this in most places. If you paid for ads to be displayed for a certain time period and they are not, then there's been a contract violation.
Provide one example that can not be solved without surveillance.
An easy fix with surveillance is to rate limit people based off their IP address. Without surveillance though there is not much you can do. Scale up your infrastructure to try and out scale the attack? Implement a global rate limit that locks regular users from being able to sign in?
Pick a more sensitive area than your IDE, say medicine targeting erectile dysfunction, sexual or religious preferences, etc. You may find that being allowed to collect that data, especially covertly, just to save some money suddenly doesn't look reasonble at all.
But surely I should be allowed to covertly collect any data about you if it enables some savings for me. After 15 comments insiting it's OK you should only approve of this.
The first thing I would do is look outside to collect information on who in outside thereby infringing their privacy.
>Also suggesting that an IP based rate limiter is the same as the surveillance in question is very disingenuous
Recording people's IPs is definitely surveillance.
>say medicine targeting erectile dysfunction, sexual or religious preferences, etc. We may be able to connect drug sellers or churches with people if we know that information.
>But surely I should be allowed to covertly collect any data about you if it enables some savings for me.
Sure you can. Go ahead.
Looking at someone doesn't infringe on their privacy. Taking a picture of that someone and storing it in a permanent fashion, might. To prevent abuse/DOS you only need to do the first (which does not constitute "surveillance" or loss of privacy), not the second.
> Recording people's IPs is definitely surveillance.
It's not surveillance if you are not tracking anything else other than IPs (i.e. no other behavioural data associated to it).
Either way, you still have not provided an example where surveillance is required to prevent abuse: I can simply store hashes of "bad IPs" (or ASNs) to blacklist... no need to store any information that could lead to an actual person (like an actual IP address).
We'll there's your problem. First you show a complete lack of understanding of the issue, from its basic concepts to the practical manifestation and consequences, and then you conclude that it must not be a real issue.
This technique can be used to justify anything. Burning books? Sure, it's like burning extra processed wood, totally okay, go right ahead.
Ignorance is not a defense.
Also can you send me your medical data and search history? I mean you're OK with sharing this data and said nothing about it being ok only if I can do it covertly. Better yet, give me your name and address and I'll just grab that myself so it's not too much of a bother for you. It's just so I can serve cheaper better targeted ads to you.
I mean refusing and backing out now would just be hypocritical and completely undermine the case you so unsuccessfully try to make wouldn't it?
This is your best argument why we have to track and profile every human on the planet around the clock?
This is totally surveillance. Just because we delete data after a while, it doesn't mean I didn't surveil you, nor does it mean I haven't used that data I got from you for my own benefit.
>This is your best argument why we have to track and profile every human on the planet around the clock? You just asked for an example. If you are suggesting that my argument is to prevent abuse of systems I would say that it justifies tracking every person on the planet.
It is not. I connected from some IP because I wanted to use your website, at the very least you have to remember my IP address for some time to send me your website back. And if I want to access your website and it will be only available if you store my IP address for a few minutes to fight off attacks, then this is a use of my IP address that I welcome because it is for my benefit. And if you really want to, just store hashes of the IP addresses [1].
Just because we delete data after a while, it doesn't mean I didn't surveil you [...]
Sure, surveillance is not defined by the amount of time you store some data. If you store my shipping address for years it is not surveillance, if you store my IP address for one second to add an entry to my record in your database that I just visited the website it might be surveillance even if you do not permanently record my IP address. But I never claimed that the amount of time you store some information is a or the relevant criterion
[...] nor does it mean I haven't used that data I got from you for my own benefit.
Also irrelevant. If you store my IP address for a short time or my shipping address for a long time in order to send me the website I requested or my order than this benefits you because you will make some profit from my order.
Relevant for whether something is surveillance or not is whether I approve what you are doing. If you track my position day and night in order to show me ads for businesses nearby it is surveillance unless I specifically requested this. If you track my position because I am using a fitness app and requested to record my run, then it is not surveillance.
[1] For IPv4 this is of course essentially pointless. But maybe you could come up with a more elaborate schema than simple hashes, maybe salt them and rotate the salt every few minutes or whatever. But you will probably not gain much besides added complexity.
Regardless, consider a DDoS attack. If every new request is coming from a different IP address, how do you continue providing service to your legitimate customers while blocking that malicious attack? Knowing the attacker's IP addresses doesn't do you any good... because they can just keep using new IP addresses, and blocking the old ones doesn't do any good.
This is where heavily surveillance-based systems like Google CAPTCHA often come into play, and I have very mixed feelings about those.
There are some non-surveillance-based captchas like this one[0] that I saw on HN awhile back, and I hope those become successful.
To complete the metaphor Amazon would use the address you gave them to help improve their business in some sense without asking you if it's okay. Similar to how web masters don't ask if it's okay if they write what pages we access into logs is okay.
>Knowing the attacker's IP addresses doesn't do you any good... because they can just keep using new IP addresses, and blocking the old ones doesn't do any good.
Then we should try to find any patterns with the traffic that we can use to try and filter it out. This is a place where fingerprinting is useful.
>friendlycaptcha
This just slows down bot spam instead of testing if someone is a bot. Someone posting spam to your site once a minute is still annoying.
I'm not going to waste my time further on this thread after making this one last point.
> This just slows down bot spam instead of testing if someone is a bot. Someone posting spam to your site once a minute is still annoying.
Google CAPTCHA is trivially bypassed all the time. Do you really think it isn't? Sometimes using services like Amazon Mechanical Turk, sometimes using simple computer vision. It doesn't test whether someone/something is a bot either... it just tests whether they can pass the CAPTCHA. It certainly doesn't test whether they're part of a DDoS, nor does it test their intentions to find whether they are good or malicious. It's just a CAPTCHA, but it also uses a lot of surveillance... and as I said, I have mixed feelings about that. I didn't mean for this to become the point of the thread, it is definitely off topic.
The idea of Proof of Work CAPTCHAs is that you can actually make it more expensive for an attacker to solve those than it would be for the attacker to solve Google CAPTCHAs. Obviously, this is still an area of debate and research.
I'm not exactly sure what this means. I used to be all for total privacy, but I found that future to not be sustainable. Perhaps I'm just jaded, but privacy just gets in the way.
>This whole thread is about surveillance based advertising being bad.
Well this part of the thread isn't. It's talking about how surveillance improves services by allowing them to deal with abuse.
>In no way is using an IP address in a firewall a form of surveillance. It isn't. The IP address isn't being associated with any other data, it's just some numbers floating in space, disconnected from any human being.
Wrong. I am using your IP as part of a scheme to fingerprint you. I want my rate limit to limit each person separately. An IP address is just a somewhat decent way to approximate that.
>The idea of Proof of Work CAPTCHAs is that you can actually make it more expensive for an attacker to solve those than it would be for the attacker to solve Google CAPTCHAs.
This has to be carefully balanced with the user experience. No user in going to want to wait 5 minutes to post when they can just have a Google account with a good reputation and just click a checkbox.
That's not your decision, I decide what matters to me, whether I want my privacy or this nebulous sustainability, whatever this is suppose to be.
Wrong. I am using your IP as part of a scheme to fingerprint you. I want my rate limit to limit each person separately. An IP address is just a somewhat decent way to approximate that.
Then let me turn this around, if using my IP address in this scenario is surveillance, then don't do it. If it is necessary, then ask me for permission, can we use your IP address to fight off attacks and ensure the availability of our website or do you prefer that the website might not always be available due to attacks? And the same applies if you want to rate limit all users, offer the choice between not using your website or opting in for IP based rate limiting. It's that easy.
The definition in the report is poor. Yes, you always need to advertise to a segment. No, you don't have to spy on users to do it.
How? Make a website about something and select advertisements that are relevant to the sort of people who are probably interested in the topic of the website. ReadTheDocs has already spun off an ad business that advertises tech stuff to readers of ReadTheDocs because it's reasonable to assume that is the audience that is perusing ReadTheDocs pages.