What's Now State of the Art in Encryption Technology?

With the events of September 11, 2001 still vividly etched into our conscious minds, it was only a matter of time before the US Government would paint the crosshairs on their next target after Bin Laden: encryption. With Ashcroft's declaration of computers as tools of terrorism, and law-enforcement pushing for enhanced surveillance, it appears that one of the first victims of America's new war may be the privacy of her citizens. Of course, if you are concerned about privacy, you're probably wondering how to improve what protections you have in place, if any. So what are the leading-edge innovations on the encryption front right now, and how easily can such tech be adapted to everyday communications? C :In an interesting display of synchronicity, Timothy posted this article, earlier today, which notes that Steganography use isn't as wide-spread as previously thought. Deagol asks: "With the Feds pushing for encryption back-doors, and even more domestic surveillance, how can we resist this? I mean in a practical way, but at the same time taking a stand for our rights to privacy and assembly. What's the current state of the art in hard disk encryption? Email encryption? Steganography? There are many tools out there, as well as many link-farms, (I looked at many today), but many pages seem dated, and it's hard to tell who's using what in a useful implementation. So, who is using PGP or GPG? Who is using BestCrypt or Loopback Encryption, Freenet or Steganography? A privacy weenie wants to know what your daily-use setup is!"

One thing about encryption: the easier it is to do, the more people there will be using it. For the non-tech user, encrypting messages on a day-to-day should be no more complex than 3 steps.

JPMH asks:"First journalists and now even relatively clued-up politicians in the UK are talking about making it an offence to use strong encryption in email and web-pages. An obvious counter is that this won't work, because the messages can easily be hidden using Steganography (Slashdot Jan 2, May 8). But that assumes that the steganography itself is good enough not to be detected. Is this true? How good is the state of the art?

To be undetectable, the properties of the 'message' bits you are putting in must be statistically indistinguishable from the 'image' bits you are overwriting. According to a paper by Neils Provos and Peter Honeyman of U. Michigan (highlighted today in the Register) the simplest common programs, such as JSteg and JPHide, fail this test badly and are easily detected. But they failed to nail any confirmed steganographic content in 2 million images on EBay.

Other programs (eg Provos's Outguess 0.2) are more sophisticated at hiding the messages (and other media eg MP3s give a bigger haystack to hide them in); but on the other hand, more sophisticated statistical models of images (eg Slashdot 16 Aug) may be better at making the 'hidden' content stand out.

So, can messages reliably be hidden? Or will people trying to hide their messages in a reliable manner get caught?"

Comments

What's state of the art? PPS.

[ajs]

Ok, I'll admit I'm biased, but I think the next phase in the developing landscape of encryption is universal access to cryptography. I'm not talking about putting PGP on FTP servers, I'm talking about making hard crypto available to my mother.

To this end, I've started the PPS, which is a project devoted to transparent, universal email encryption. The goals are complex, since they are aimed at so many audiences, but you can browse the site and get an idea. If you find it to your liking, please drop me a line and sign up to help.

You don't have to have technical skills. I need proof-readers, coders, researchers, and more. The reference code is not nearly as important as getting the specification done and doing all of the research needed to get the various MUA vendors to sign on.

Re:Tools of Terrorism

[monkeydo]

You make a very intersting point that will no doubt be lost on most of the Slashdot audience (as well as yourself I suppose)

Airplanes;
Dynamite;
Plastic Explosives;
Fertilizer chemicals;
Telephones and other communication equipment;
Knives; and
Boxcutters


Are all heavily regulated already. Some directly like explosives and airplanes, and others indirectly like phones and knives.

Why should strong encryption be different? Just about any tool you can think of has good uses and bad uses. That doesn't mean we should ban the tools, but we should try to minimize their use for purposes contrary to the common good.

Does it violate some inalienable right that you cannot walk into walmart and by C-4 off the shelf? Certainly you have some harmless use for it. Should convicted felons be allowed to carry firearms on the street?

Wake up to the real world people. The fact that we live in a society means that we voluntarily give up certain freedoms for the common good. That is the decision that groups of people make when they get together and form governing bodies.

You cannot simple say banning==bad freedom==good unless your definition of good is anarchy. Do we all agree that the ban on murder is good? Even though it takes away my right to express myself with creative killing?

Getting steg to work

[iabervon]

First, share a one-time pad. This is very easy using steganography: you just choose an image on the internet and a time and agree to seed a pseudo-random number generator with that to get your pad. Encrypt your message by XORing it with the one-time pad. Your encrypted message is now indistinguishable from random noise, assuming your PRNG is good.

Then, you need a data file where noise is expected. Using low-order bits is no good unless you have pictures where the low order bits are actually random, rather than containing no information. One possibility is to take a photograph and make it a GIF or PNG; the lowest order bits that your camera actually produces are probably noise, and will be present in the image.

Replace the input noise with your special noise. The resulting image is now perfectly plausible (your camera could have taken it if some photons happened to land differently, with the same probability as having taken the photo it did take), and the message cannot be read or distinguished from noise unless the codebreaker knows what image you agreed on.

In order to do this, you and the recipient have to agree on an image you control and another image. Having done this, you can, of course, agree on more images later, for communications in both directions. Make sure you both look at a lot of images, including a lot that everyone looks at (e.g., CNN).

And then your recipient looks at the message on his CRT, and the spies read it in the EM radiation. Good thing you weren't saying anything they care about, but why did you bother with all the encryption, then?

You have no chance to decrypt, make your time.

[ZanshinWedge]

Seriously though, if you are highly technically savvy (which I will assume since we are speaking about the state of the art) then you can not only create near unbreakable encryption, but near undetectable (or untraceable) encryption. Steganography is a child's toy compared to some of the things that are possible. The internet is a vast 86,400 / 365 information sea, slipping a few megabytes of low profile data into it is going to be hard to notice. By utilizing multiple techniques at the same time (hard encryption, low signal to noise ratio channels, low detectability communications, difficult traceability, etc.) you can be confident that even if someone found your data they would not be able to understand it or extract useful information from it.

For example, let's say you want to send data to someone else. Let's say it's a short text message, though it could be anything up to gigabytes of data without too much trouble. The sender encrypts the text using public key cryptography with a large key (4096-bits or larger), then breaks the encrypted message into several really small chunks, then uses a program to generate thousands of fake chunks. Then, using a sequence of hacked ISP and shell accounts (preferably spanning the world), the sender embeds this "chunk stream" into some nondescript form of communication. Let's say they use a large number of spam messages, or pornographic multimedia posted to a highly trafficked usenet newsgroup over several days and a simple steganographic technique for the embeddding. The receiver downloads the source files, extracts the "chunk stream", selects out the valid chunks, then decrypts the data.

Let's say that Los Federales were able to detect that something funky was going on. That alone, in the firehose of the internet, is a significant challenge. They would need to first be able to extract the data from the embedding system. Not impossible, but difficult. Next they would need to cull out the invalid chunks in the pile they now have. This can be made as difficult a problem as breaking hard-encryption in and of itself. If they manage to wade through that mountain of sludge, they end up faced with near unbreakable encryption. For added fun, repeat some of the steps multiple times! (for example, double encryption, double stage steganography, etc.), preferably with different techniques for each iteration (encryption cycle 1 uses RSA, while cycle 2 uses elliptic curves, etc.)

Or, you could take the route the US has taken since before WWII and use one time pads. One time pads are provably cryptographically secure (if you don't have the key you simply CAN'T break the encryption). The only difficulty is distributing the keys.

Nevertheless, I would imagine that the main goal these days would be low-detectability rather than pure cryptographic security. If they can't find your pigeon in a flock of wild birds then they very well can't even try to decrypt the message it carries. There is a LOT of noise on the internet, that provides a huge amount of hiding space.