Furthermore, they wouldn't have to worry about deleting customer data either. The customer would have the power to simply deny access to the keys.
But Google needs to know my search query to be able to serve me a results page. Google will know my phones location if I use their WiFi-positioning-system. Google needs to know my address and payment information to have me buy apps.
Encryption can not help me here. I need to be able to trust Google not to collect that data without my consent, that they do not store that data longer than needed to provide me the service and that they do not share that data with anybody that has no right to it. That can only be guaranteed by laws.
This is about laws, and the need for such laws will not go away by encryption alone.
Otherwise, there'd be plenty of services competing with Drive, DropBox, etc, that did just that.
There is no other way. Unfortunately, there has been a serious lack of r&d in this area, so we have a lot of catching up to do. I believe it can be made to work, though, because this is not entirely a new idea for most people: they already understand physical keys and the problems associated with losing them.
Moving to digital (public-)keys isn't a perfect match, but it is entirely possible to leverage the skills and experience that people already have. For example, contrary to the advice used a decade ago, Schneier recommends[1] that people write down their passwords (avoiding the memorization problem). This can provide good enough security most of the time, because people already understand how to provide security for their wallet.
Incidentally, I always liked the basic idea of the Java Ring[2]. It was basically a fancy (for the time) smart card, but the idea of storing the chip in a ring is an attempt to make security management personal, which is an idea that had potential.
[1] https://www.schneier.com/blog/archives/2005/06/write_down_yo...
[2] http://www.javaworld.com/article/2076641/learn-java/an-intro...
edit: forgot a link
You can't do this with digital keys because a digital key that is weak enough for this strategy to be usable is also too weak to protect you from the main class of attacks it's supposed to be protecting you from.
In the digital world you have Shamir's Secret Sharing. Generate a non-encrypting, non-signing, decode-only key and give a N-th of if to N escrow services. If you lose your key you contact M of those entities and recover the data.
Plenty of tricks have been created in the last two decades to address the problems associated with key ownership, the problem is always the lack of UI and of standardization in general.
Store your N'th fragments with /friends and family/ that you know well. You might also be able to get in on distributed backups together.
If the stuff being protected by the key is, for example, the type of data that someone would otherwise keep in a personal fire-safe, then simply printing out the keys and keeping them in that fire-safe wouldn't change the type of security being provided.
It might be nice if we had some type of highly-reliable (like the redundancy in QR codes[1]), so the process of printing out a key and re-scanning it was easy and trustworthy.
Well, we can always legislate. Even with encryption of data, you want legislation. Without legislation, the Government will just create a separate channel for data collection to bypass your encryption. It starts unencrypted at the remote site, so it's not that hard to do.
Even with legislation there will be problems, and overstepping, but at least then there's some semblance of recourse and ways to fight. You can't fix the problem for everyone and always, but you can fix it for most people most the time. This is the sort of thing that should be in trade agreements and treaties.
Depending on your target market, that can be a selling point. Tell them early and tell them often that you don't keep copies of their keys and if they lose them, they won't be able to access the information. Remind them to make backups.
Outlook works just peachy searching my GB's of emails and attachments from its local replica, using the internet only to sync.
No one will use Gmail if they need to download GB's of Junk they haven't cleared in 10 years to find an email for 2 months ago, not to mention that any type of sharing will force major duplication as well as require you to download and generate new copies with each person you want to share it with.
[1]: https://developer.mozilla.org/en-US/docs/Web/API/IndexedDB_A...
And I don't want to waste power and time syncing all that data between those devices just so i can do the same things i can now but less efficient (and with still more drawbacks)
(inb4 "server-side amazon search doesn't work anyway" ;)
Spam filtering can be done as the email streams in. Email databases which power the spam filter can be de-anonymize do and reduced to statistics.
I guess sharing is harder but it can still be done. If I share content with you, I'm giving you permission to have a copy so it can be encrypted with your key as well as mine.
Sharing a key involves downloading it from the server, asymetrically encrypting it with the public key of the reciever, and uploading to the server again. It also allowed sharing with groups via intermediate keys decrypt-able by multiple users.
Obviously it's just a PoC - noone should use it for actually important data, and I certainly don't promise that it's actually secure - but it does (more or less) work. As it requires client-side crypto, it's not an ideal candidate for a web-service either (although one could potentially use browser extensions to make that more reasonable).
What about Hashicorp's Vault?
Admittedly, there are some recommendations on the Tarsnap site (http://www.tarsnap.com/gettingstarted.html):
"5. Keep your key file safe
Store your /root/tarsnap.key somewhere safe. If you lose tarsnap.key, you will not be able to access your archived data. The same encryption which ensures your data security also means that there is no way for anybody (including Tarsnap Backup Inc.) to restore your data without this key. There are many ways to keep it safe: Copy it to a different system. Put it on a USB disk. Give it to a friend. Print it out (it is printable text). Store it in a bank vault. Pick at least one and do it!"
The only other way I can think of to keep the key even safer would be to use Shamir's Secret Sharing to divide the key up and share it among trusted friends and family. (https://en.wikipedia.org/wiki/Shamir%27s_Secret_Sharing).
How would users verify that the software was working securely? A webmail provider would need to publish a full client-server API and protocol spec, so that you could run some independent software to audit what was being sent to be sure no leakage was happening. This is simply untenable.
Oh and what would really happen is that everyone would lose access to their data since, in general, nobody is able to maintain keys. It's pretty understandable that companies have little desire to redesign their whole systems to provide a crappy experience to a nearly non-existent market segment.
Think of a Cow in a barn. The food/water/shelter is free, because you are what's to be sold.
I am the only one that is aware of symbiosis?
Still arguing that life in Googles and lately Microsofts barn is quite comfortable and a whole lot a safer than outside because those companies has a huge interest on securing their users
Bitcoin decentralized money.
Until now, social has not been decentralized effectively.
Security is even harder, perhaps! Because it requires authentication but also a clear definition of guarantees, including availability! Perhaps your data could be stored on your computer and other replicas can be stored encrypted in the network, like freenet, but encrypted only by you. And your key would be a hash of your passphrase like "all horses are fine". If that guy's bitcoin key had been a result of a hashed passphrase, he wouldn't have lost $3 million in the garbage!
It exists in 3 forms.
1) Both benefit (I.E.: mutualistic)
2) Only 1 benefits while the other has no clue of its usage (I.E.: commensalism). Vultures and Lions more or less, or hermit crabs and shell fish which are dead.
3) One actively benefits while the other is consumed by the active particpant (I.E.: Paracitism).
:.:.:
Your argument is that webscale logging is mutualistic. To an extent I'll agree. But this isn't the case. Google is mutualistic when using say GWallet, Docs, Mail, search etc.
But Like/Plus buttons, as well as cookie sniffing isn't mutualistic. Its commensalism. I have no say if another organism uses what I leave behind, it doesn't directly effect me, but it benefits another organism.
Google/Facebook/etc are not selling your data, they are not selling you, they are selling access to you via advertisers, more specifically, they are selling access to people who like certain things and who live in certain areas who have certain demographic profiles.
If you're curious about this, go sign up for an Adwords account and see how the targeting works - what you see there is the limit of what any advertiser using Google sees about its users. Facebook is much the same.
Get it right. Any ad-backed company of note does not "sell your data" and characterizing it as such is so divorced from reality as to be a blatant, willful lie.
And since they're also not allowed to give the data away and can get into trouble for allowing it to be stolen or accidentally shared, my name and personal details are also somewhat less likely to crop up in an embarrassing public "people who subscribe to this service" list...
Nothing in the ad-backed company model prevent someone sending a fake advertisement to the demographic who are likely involved in adultery. After that you publish a "people who subscribe to this service are ..." list based on who clicked on it.
The only question if someone could do this with Facebook/Google is if they provides targeted advertisement to that demographic or if you have to provide your own algorithm and combine different targeted advertisements with the available demographics.
The false advertisement might say something like "save 50% of your next amazon purchase" or "try out our new car service by getting the first trip free" or any other ways of getting people to voluntarily filling in information which looks completely innocent. All you need is the additional information of name and address, which for the user feels completely separate from the sensitive information which was surreptitiously inferred through the targeted advertisement.
This assuming they can't simply push some malware and get the name that way. We constantly hear about tor hidden service attacks done by de-annonymize people with flash/java/browser exploits.
Do you not realise how low response rates to ads and promotional mailshots are? Try as you may, you cannot possibly hope to obtain anything more than a tiny fraction of the data held by a third party by advertising your phishing page on their network, even if the third party actively supports your goals. And even if you get some useful personal information, if you publish it without consent, you've committed a crime.
Banning the sharing/selling of data without consent imposes significant obstacles to legal and illegal hassling of people whose data has been mined.
Precisely for some of the reasons outlined in the above article (third party records are subject to subpoena pretty much everywhere, not just the U.S.) it's untenable in certain fields to keep sensitive information on the cloud.
Those concerns would be dramatically mitigated if at least there was an option to do client-side encryption. That way the government has to get jurisdiction over your person to get you to turn over the encryption keys, which puts you on much stronger legal footing.
Siri needs collective anonymous data to work and improve itself. Ads, like them or not, cannot function without large collections of data. No attempt to predict customer behaviour is possible, no usage analytics, no cutting edge improvements, a lot of cool things wouldn't be possible.
It's just how it is.
There is no value to the company in doing that. They can't mine the communications to determine what ads to serve you, or anything else. At that point they'd just be offering a free service. Why?
So many people are saying "your proposal breaks cloud services or requires client side compute", but that's not true. Be more imaginative and get outside the current model.
From a somewhat elitist perspective I actually prefer the status quo. At least, today, we have backup services like SpiderOak or tarsnap and we can use encrypted email (I don't) or other communications software.