56

In certain jurisdictions, use of cryptography by the private sector is limited: e.g. there are reports that in the UAE and other countries not all of the encryption capabilities of the BlackBerry are permitted. Similarly, in the 1990s the U.S. government tried to force the use of Clipper, encryption hardware that would have led to all private sector encryption keys being escrowed.

Let's assume that your business needs to operate in a region where encryption is either completely forbidden, or its use is limited to the point where you don't believe it is effective. Of course, your organisation's confidentiality requirements haven't changed...

What do you do? Do you believe it is still possible to protect confidential data? If so, how?

Update as different sovereignties have different ways of writing, amending, enacting and enforcing laws, it's hard to say exactly what the text of such legislation would look like. Feel free to make (and state) assumptions, within the scope of the scenario described above: that encryption is not legal or not useful.

8
  • 3
    The device with automatic key escrow was called "clipper". Skipjack is the name of a block cipher which clipper used, but that's not the core of the escrowing system (Skipjack is a relatively decent block cipher -- albeit with 64-bit blocks and 80-bit keys -- designed to be efficient on the kind of hardware that clipper used). Apr 19, 2011 at 11:32
  • 3
    Tangent: Does anyone remember the case with a boy who was barred from being able to use encryption (along with some other things) by a court? This is just absurd techdirt.com/articles/20101020/04513511498/…
    – Ormis
    Apr 19, 2011 at 14:19
  • 1
    Its easy by a safe and put your data in it. If you want to send data to someone else by a smal portable safe put whatever data you what to send on a portable usb and putt it in the small safe and UPS it! Problem solve i cal it IRL crypto ;) Apr 24, 2011 at 21:43
  • 1
    @Ormis, I assume you mean steganography? Although some stenagrophists could really be doing that too...
    – AviD
    Apr 26, 2011 at 19:26
  • 1
    Countries restricting cryptography that much will not have any remorse to force you to reveal those informations even if you want/can hide them. Trying to hide information could reveal more risky than "hiding in plain view" as any discovery of you deliberately hiding information from them will suffice to close your business in a matter of minutes, don't take that risk and keep out those countries.
    – Shadok
    Aug 18, 2011 at 14:02

17 Answers 17

21
+200

Thanks for the insightful question. The more I think about it the more it feels like someone has pulled the rag below my feet (living without cryptographic protection).

Analyzing the resulting threats (strictly from the business point of view - the requirements of dissidents etc. are a different story), I see them being:

  1. Corrupt government officials who will (ab)use their position to sell confidential enterprise information passing through channels they control.

  2. Business antagonists who will try to eavesdrop on your communication channels.

A very first thought would be to work on two fronts; diversification of paths and leak identification:

Without (or with watered down) encryption I would think that one should try a. to break down critical messages to smaller chunks and then b. diversify the paths each one would take. He would thus make life a bit more difficult to people eavesdropping by multiplying the paths they would have to work with. I suppose some of these paths should be more "traditional", like normal post, messengers etc. that the government can inspect but not the antagonists.

This of course does not exclude the government from reading all messages from all channels simultaneously, so I would suggest a kind of watermarking solution to identify "non authorized leaks" coming from corrupt government officials. Or perhaps also adding fake (but traceable) messages in the mix?

Even governments that for their own reasons want to inspect all communications don't want to be branded as "business unfriendly". They would then probably take steps to punish any of their officials that abuse the eavesdropping capability for personal gain. A few, well publicized, cases of gov officials caught red-handed or just an announcement that the enterprise can effectively track leaks would help prevention big-time.

Now, how about a solution for the "always-on" blackberry crowds?

2
  • Yup, the crackberry is one of the more important problems I envisage. Corruption of the escrow mechanism is a very interesting attack scenario.
    – user185
    Apr 19, 2011 at 16:16
  • The thing is though, they do not labour under the same restriction. They can use 'the cloud' all they want and that is already seeded with a great deal of our data.
    – chiggsy
    Apr 24, 2011 at 20:31
18

In theory you should still be able to achieve confidentiality protection in some circumstances, because crypto isn't the only way to provide confidentiality, you can also provide it via access control.

Realistically however it is difficult to think of any real world system where you can usefully achieve this without at least something like SSL in the toolbox, and the governments may insist on being able to bypass access control too. Also, distributed access control generally relies on some form of crypto, though governments that restrict encryption sometimes make exceptions for that.

Also I guess if you can accept a threat model where the governments (and only the government) can bypass your controls, then it may still be worthwhile implementing controls to protect against other threats.

Last point is you can possibly look to implement a system whereby if the government accesses client data, you at least know about it.

Update: as a side point, it is perhaps worth mentioning that it is useful to separate out the following concerns:

  1. crypto used for privacy

  2. crypto used for integrity protection

  3. crypto used for authentication of principals

A number of reasons for this, firstly historically laws have made a distinction between these use cases in some jurisdictions - so, e.g., crypto for privacy not allowed, crypto for integrity/authentication OK. Secondly, 2&3 are useful on their own (e.g. for distributed access control). Thirdly, some protocols use a different key for integrity and confidentiality, or can apply integrity protection independently of confidentiality - so it is not always necessary to weaken integrity protection and authentication just because a government insists you weaken privacy.

13
  • It'd be interesting to consider whether 'portable' data (smartphones, laptops) can still be used if access control is the only secrecy mechanism you've got.
    – user185
    Apr 19, 2011 at 16:14
  • If you'd got a way to wipe the device (or not persist data on it in the first place) then perhaps. Quite often what is billed as 'encryption' on these devices is really just a quick way to wipe the device anyhow (i.e. erase the key instead of the memory). Apr 19, 2011 at 22:21
  • Reliably wiping smartphone-esque devices currently requires encryption...
    – user185
    Apr 20, 2011 at 4:46
  • True, but this is more an SSD/flash implementation bug/peculiarity than something inherent in the data being portable. A laptop with vanilla HD might be acceptable if coupled with a remote wipe scheme for example. And not persisting the data on the portable device at all is sometimes an option, albeit probably at the expense of performance and convenience generally. Apr 20, 2011 at 11:26
  • @frankodwyer: sure, it's a peculiarity of the way smartphones work, but "change the way all smartphones work" is quite an expensive solution.
    – user185
    Apr 22, 2011 at 10:10
16

I think there's not enough information to answer. Without seeing the exact text of the law, we can't say whether it will be possible to communicate securely; it depends. That said, I will make two small points.

  1. I think if legislators really want to prevent secure communication, they'll do that. Sometimes technical folks think that legislators are dumb and will write something narrow, like "every encryption algorithm is prohibited", so all that's needed is a clever technical scheme that doesn't count as an encryption algorithm. Perhaps, but I think in the long run, that's not a great mental model. I think if legislators want to outlaw all methods of communicating securely, they'll write the law broadly.

    For instance, if government is going to ban crypto, it's probably also going to demand access to your data, and leave it up to you how you are going to provide access. (Consider CALEA as a model. The US government didn't write a law that says "encryption is illegal". Instead, they wrote a law that said "telecommunications providers have to give a copy of their communications to the FBI whenever the FBI asks", and left it up to the telecomm providers how to achieve that. One consequence is that this restricts use of crypto, but it also restricts use of other alternatives that some folks here are suggesting.

  2. There are alternative ways of protecting confidentiality, which don't look like usual encryption. See Ron Rivest's chaffing and winnowing for one example. That said, in the long run (and possibly in the short run too), if legislators want to ban encryption, they'll probably ban substitutes, too.

2
  • +1, for exposing a premise: The relationship between the legal definition of the term "Cryptography" and the term used in computer science is not clear. Apr 30, 2011 at 17:33
  • There are other ways to protect communications that would be difficult to ban. There was a sci-fi story (sorry no cite) about people paying a company to publish huge amounts of slanderous material about them to make it impossible for investigators to separate the fact from the fiction. As long as eavesdroppers don't have standing for slander against you, it is possible. Large numbers of fake but equally plausible messages can leave an attacker guessing but the intended recipient can use a secret to find the real one. No crypto, but a shared secret can still establish a secure channel. Apr 30, 2011 at 17:39
13

Even without cryptography, you could use steganography to make information hard or in some cases near impossible to find. It seems difficult to imagine a situation under which government could effectively ban that, since one could claim that a secret message was hidden anywhere and it is very difficult for an accused individual to demonstrate otherwise (essentially this boils down to the impossibility in law of proving a negative).

Government could, of course, ban steganography too, though I think it would be very hard to enforce.

3
  • +1: I agree that steganography would be hard to ban, and further that it does at least partially hide data "in plain sight". I wonder where all of the necessary "fake" information that the real info is hidden in will come from, though...
    – user185
    Apr 19, 2011 at 16:18
  • 2
    @Graham Lee - the "fake" data would consist of other people's real data. The concept was proposed by Rivest (of RSA fame) several years ago. For details, see people.csail.mit.edu/rivest/Chaffing.txt Apr 20, 2011 at 4:13
  • Steganography cannot possibly be banned without utterly destabilizing society. Not that it hasn't been tried. There are many accounts of people throughout history having been banished or executed for having said something they thought was circuitously cryptic enough but actually wasn't, or for something innocuous that was suspected of bearing a concealed meaning. Stories of this sort of excess are one of the commonest examples of heavy-handedness in government. Apr 25, 2011 at 2:38
9

How about the radical approach "Stop storing information about your customers"?

This is only workable for simple use cases, where the cost of having the data outweighs the value of that data.

Obviously this is a pretty important goal for this 'hypothetical', and if you must do business in a country, then stop storing every bit of information about a customer.

  • Make your customers anonymous.
  • Throw away the info once it has served its purpose.
  • Create your information without cross linkages between datasets.
  • Hash information (if the law allows) to limit the uses of it.
  • Forget about making easy to use/social web-sites (i.e. amazon one click is out)
  • Create plausible deniability about the information you do retain (probably needs it's own post).

I know that this is more likely to get me branded an anti-capitalist, but the most effective way to stop governments (and [cr|h]ackers) snooping is to throw away the information.

EDITED TO ADD: This is a serious attempt at a solution to the problem, examples of this are very common.

  • Illegal goods and transactions use cash means that you don't have info on your customers (and them on you).
  • The classic 'cell' approach to organising a resistance or terrorist organisation.
  • Anonymous membership in a group: KuKluxKlan wear masks.
  • Plausible Deniability: The Mob laundering money through casinos.
1
  • Good answer! Fast-forward to 2017, and Europe's GDPR is essentially making exactly this a legal requirement, with significant punishments for non-compliance, and it applies to anyone, anywhere in the world that stores personal data on EU citizens. Privacy by design!
    – Synchro
    Aug 8, 2017 at 13:03
8

Between cleartext and cryptographic encryption there's a vast area of obfuscation. From natural stuff (a foreign language/alphabet), through technical (little-endian, UTF-16, EBSDIC), through oddities/obscurity (large binary files with internal, unpublished structures), to non-cryptographic-grade encryption (ROT13). They can all be defeated with enough time and resources, but if the data itself isn't worth much (ie. "I'll be 15 mins late to dinner" emails), then such measures just might be enough to prevent the crimes of opportunity (WiFi sniffing, lost laptops).

2
  • +1 for putting into context of risk and value, i.e. sometimes this might be the right solution. Though of course this shouldn't be relied on when you really need the confidentiality.
    – AviD
    Apr 20, 2011 at 18:50
  • 1
    Nitpick, but rot13 is encoding, not encryption.
    – forest
    Nov 15, 2018 at 22:20
8

About the UAE and Blackberry: as Schneier puts it, "this is a weird story for several reasons". The UAE are involved in a complex negotiation with RIM, but, in the same time, they do not ban Gmail, which is HTTPS-only and thus fully encrypted and outside of reach of UAE law enforcement agencies (unless the UAE made some secret deal with Google, of course). Encryption looks more like a pretext than a primary motive here.

For your problem, you must define the attack model with a bit more precision. Is the government an enemy ? The point of anti-crypto regulations is to allow eavesdropping by government agencies. So you have to decide whether you mind that or not.

If not, then the scenario is the following: you want to protect your data against competitors, but within the limits permitted by local law. So what you are after is a deal with the local police forces, such that you may encrypt data but with a key escrowing system. Chances are that they already have a handy software solution (just like the Clipper chip was meant to be a hardware solution).

If the local government is an enemy (e.g. you strongly suspect that the local government will try to give a copy of your trade secret to a state-sponsored competitor), then things are more difficult. You have the choice:

  1. Either you break local law and use cryptography, but you do it in a covert way such that the police will not notice anything. This road leads to steganography. In a business context, this looks hard to use, because you would have to transmit your confidential data hidden within innocent-looking traffic, which in turns requires a legitimate reason for having so much traffic which does not look like normal data flow for your business.

  2. Or you decide to stay within the bounds of local law, but with irksome tricks to discourage eavesdropping. Among the possible tricks are:

    • Implement custom transmission protocols, not following any published standard; in particular, use compression (data compression is a very good excuse for making things wholly unreadable). The recurrent complaints of opensource advocates against closed formats show that reverse-engineering is not easy -- and you can change the protocol every now and then for perfectly valid reasons ("optimization": a magic word).
    • Encryption usually relates to a "secret convention": the data is unreadable to anybody who is not savvy to some specific convention which is kept secret (in cryptography, that convention is normally embodied by a key). So you could try to use a "discreet convention": a convention which is nominally public, but which the eavesdropper will have a hard time learning nonetheless. As an extreme example, instruct your staff to speak over the phone only in Navajo (slightly less exotic languages like Dutch or Finnish will usually do the trick).
    • Use physical protection whenever possible. This may entail some drastic measures: for instance, instead of transmitting data over some network, organize for regular sending of hard disks. As far as networks go, a suitcase full of hard disks has terrible latency but excellent bandwidth.
    • Make data go through several nested law-compatible systems. For instance, use Gmail. Even if the local police knows how to decrypt Gmail data (because of some negotiated escrow system, if such a system exists), this will make eavesdropping more complex, hence more expensive, hence less likely to be applied.

    All this relies on the idea that if the law bans cryptography so that police forces may spy on people and businesses, they are rarely ready to admit it in plain words; so they will not openly complain about how your "optimizations" make their job harder.

It shall be noted that not all countries have laws on encryption; see this page for a survey on the subject. For instance, there is no crypto-banning law in Syria (which is kind of surprising, given the political situation in that country).

1
  • 2
    Are you sure that browsers installed with OEM equipment sold in the UAE don't include and additional CA that would solve their gmail spying nicely with a good old MITM attack? That's how oppressive Tunisia did it with help of good old Microsoft, I doubt they were alone. There are way too many CA trusted by default, I can only assume quite a few are spook controlled... See news.ycombinator.com/item?id=2138565 for how Tunisia spyed on gmail Apr 27, 2011 at 1:28
8

Though I much prefer relying on Access Control (if any form of crypto is out), and would even consider steganography as a possiblity, and my gut is telling me "NOT really possible"... but since you asked for different types of answers...

And I assume that @Andrew's excellent solution of not keeping the data in the first place is not an option, since that would provide Confidentiality, but at the expense of Availability (and Integrity, if you consider a modified version of overwriting with random data)... And one leg of the CIA triad is not much use.

I also assume necrography is not an option for you (I just made that up - but "Dead men tell no tales"... ;) )


But for a real answer, I would suggest that another option is split knowledge.

If you only know half the data, split in such a way to be essentially useless, you cannot reveal what you do not know.
There is still the risk of collusion, or since we're talking government they may nab both dataholders, but this will still lower the risk exponentially. And of course it can be split in more than 2 ways...

But there is still risk when these pieces of data are combined, of course.

1
  • 1
    Wish you could have +1 for split knowledge and another +1 for necrography :-D
    – user185
    Apr 29, 2011 at 19:27
7

I'll answer this in a general way. I have spent a lot of time in China in a highly visible but actually quite unimportant role, and have had to think a lot about this issue. I'll assume you are interested merely in legitimate privacy rather than breaking the law in a morally serious way. You don't want to get in trouble for using illegal means of keeping information private, but you also don't want to give up your privacy just because it's inconvenient to maintain.

If there is a single key idea I've found useful it's to combine being natural, normal, and attentive in everything one does in the sensitive environment. That means that whatever requires confidentiality has to be only a small proportion of everything you do, so that one's activities can be normal and seem normal to an observer. If you need a lot of confidentiality, you must either generate a really huge amount of distractingly "normal" activity within which to conceal it, or else it would be simplest to find somewhere else to live and work.

A second point: there is a kind of dance that takes place between the privacy-loving citizen and the enforcers of privacy restrictions - when one recedes, the other advances, and then back again at intervals. That means you can count on periods when no one will pay any attention to what you're doing, and if you can wait till those times to do things that might be considered sensitive, you will not be noticed. Of course, figuring out what those times are is not always so easy...

6

Here's my £2e-2 so far (though please don't stop providing answers, there's plenty of interesting discussion going on):

  • Third-party 'cloud' hosting would become unpopular. I know that those people cannot be encrypting my data, and I only have their word that they won't - deliberately or inadvertently - allow their employees / the government / my competitors access to it otherwise. This may sound familiar.
  • First-party server hosting would be more popular, because organisations will want to avoid local file storage on smartphones, laptops and tablets. The default behaviour for a tool will be for all data to be hosted on the server, with a native app at the (managed) client end for the user interface.
  • If there's another jurisdiction where encryption is permitted, companies will host there as far as local legislation permits. However, for this question I'm willing to assume that the local government won't allow that workaround.
  • The reason for requiring a native app on the client is that if all company data were accessed via the web, then spear phishing would become much easier. With native apps then the attackers would need to observe the traffic or deploy a Trojan horse; the latter is made harder because the client devices are managed.
  • Companies would have policies over what information can be discussed on mail/teleconferencing vs. face to face meetings (as they probably already should have).
  • Working from the coffee shop will cease to exist. Of course, in most places wired LANs will replace wi-fi: integrating smartphones is left as an exercise for the reader.

The various obfuscation/steganography answers reduce the chance that an attacker finds something when they do see your data, but they seem to me to be taking the position "I would still like to believe that I have encryption". It would be bad to rely on it as a data protection mechanism.

2
  • Third-party storage would be useful, so long as you have a way to encrypt your data on your side before shipping it off. If you trust the strength of your encryption you could even store it in plain sight.
    – l0b0
    Apr 28, 2011 at 7:23
  • 2
    The question is about not being able to use encryption.
    – user185
    Apr 28, 2011 at 7:42
6

I wonder to what extent one could use cryptography such that those viewing it wouldn't know it was encrypted data? So it's sort of encrypt and then use steganography on that so that it looks like real data as is, but underneath is encrypted data?

Geek took an encryption class with Johns Hopkins Center for Talented Youth (a great class if you have a kid interested in such things) and for his final project he hid text in music (steganography) that actually sounded like real music. I wonder if he could have encrypted the text first and then hidden it in music.

So you'd be using the steganography to hide the fact that you were using forbidden encryption.

1
  • I like this :-)
    – Rory Alsop
    Apr 25, 2011 at 9:07
6

If you can show us the text of law that would be best. My answer would be (and by this answer, I am keeping it simple since I don't know your scenarii) INCREASE NOISE on whatever you are sending or storing.

That would mean you would end up by making them consume huge among of resources, VALID resources/data that is simply noise. Off course You should be aware of a way to identify your data quickly and the valid ones.

Since, I assume they are forbidding encryption in order to monitor the traffic and your storage, by having a lot of valid noise they would end up with a high ratio of FALSE POSITIVE that would drive their analysis almost useless.

Then the cherry on the cake would be to protect your very special data with a camouflage way, data in images, images in sound, ... you add some steganography to kill the monitoring processes.

This solution would cost a lot of additional overhead but would BE the best shot in my eyes given the constraints you are adding.

After writing what I have wrote, now I am thinking: Security by Obscurity sometimes makes sense ;)

3
  • 1
    I'm struck by how several of the answers say essentially the same thing, which I think you've described here most clearly. But Rivest's Winnowing and Chaffing is essentially a way to hide stuff amongst lots of noise. And texmad's answer says the same thing in more human terms. And of course steganography is another technique for doing that.
    – nealmcb
    Apr 25, 2011 at 15:20
  • The problem is that there aren't a magical solution. If you want to abide by the law (severe law in this case) you would hide your customer's confidential data. Still IMHO if the government imposed such restrictions, nor steganography, nor noise nor anything would stop it from getting it. I mean IT Security is one thing but when they want your data they would know other means (and remember they create their law) to get to them. Apr 25, 2011 at 15:46
  • Here is a link to Winnowing and Chaffing for interested readers people.csail.mit.edu/rivest/Chaffing.txt Apr 25, 2011 at 15:48
5

I'm not sure what level of protection you require, I think that in this situation I would probably consider hosting my confidential data outside of the jurisdiction of the blocked country and sort out VPN connections back to the data as required. Use of OpenVPN or similar open source apps would give you some confidence that the VPN software you are using has not been compromised.

It won't work for everyone but might be an acceptable solution.

3
  • 2
    Wonder if the crypto restrictions includes breaking the SSL for your VPN :-)
    – Rory Alsop
    Apr 19, 2011 at 11:10
  • 2
    Yeah, I agree with @Rory, that if encryption is outlawed you can't rely on encryption as the solution.
    – user185
    Apr 19, 2011 at 16:16
  • This could work for a few very narrow situations, such as securely storing medical records for a group of travelers including a medical doctor. In an emergency, the doctor could get access over an unencrypted link to the injured traveler's relevant information -- even though everyone knows that eavesdroppers will also be able to overhear that information -- without revealing any of the other traveler's information.
    – David Cary
    Oct 19, 2015 at 14:32
4

One that is currently used where traditional encryption is impossible (often because the fact of encrypted comms would divulge identies) is one time pads for entire phrases. I'm obviously thinking of spies:

"The Old Swan Seeks A New Home For Winter"

could mean anything you want - this would require pre-arrangement of phrases, and is only really suitable for occasional high value messages from a small list, but it works.

Probably not fully true to the spirit of the question though :-)

3

This is a very thought provoking question and everyone ahead of me seems to have said everything I would have said, plus a number of ideas I would have never thought of.

Most of the answers come to the same answer: stenography. It makes me think of stenography from a different angle. What if you view the message as hidden from even your receiver and the receiver is doing a side-channel attack on your noise+message? Some messages leak information for someone knowing what to look for.

Your laptop logging data everyday seems like a lot of innocuous traffic that could report home every day, slightly modulated.

2

One method of protecting data in transit could be, as hinted at by @dad, encrypting a steganographic message in the ISN number of the TCP header, following all the correct protocols. Firstly this would be very difficult to detect unless you have a warden screening all the ISN numbers passing through a server. Secondly, as the ISN is randomly generated any warden would have trouble detecting encrypted data. And of course as it is encrypted the data would be relatively secure should it be discovered.

This is discussed in a lot more detail and clarity here http://team.gray-world.net/papers/ih05coverttcp.pdf

1

The best solution is not doing business in that country. Without secure encryption, sensitive data can not reliably be protected.

If that's unworkable, do whatever it takes to keep sensitive data away from there.

Using illegal encryption is a good way to lose your business and freedom.

You must log in to answer this question.