Slashdot Mirror
New Software Secures Data when Owners Walk Away
Makarand writes "Leave an operating laptop unattended on your desk and your sensitive data
is accessible to anyone who gets hold of it. To limit this risk many users
configure their systems to fall into a "sleep" mode after a period of inactivity
and ask for a password before the system can be awakened. This constant re-authentication
proves to be a headache for many users. Now a Professor and his
graduate student at at the University of Michigan have come up with a system
called
Zero-Interaction Authentication (ZIA),
described in this article in The Age,
to protect data on mobile devices.
The system works by starting to encrypt data
the moment the owner walks away from the system. The owners wear a token with
a encrypted wireless link with the laptop. If the token moves out of range the ZIA
re-encrypts all data within 5 seconds.
If the cryptographic token moves within range the system decrypts the information for the
owner.
The token, which could take many forms, is currently a wristwatch with a processor
running Linux designed by IBM."
would it not be more sensible to make the token a passive device, like one with an RFID
I'm not an expert in encryption, but I have had serveral security related dongles and all of them were a pain in the arse.
it would seem that there are technologies (I've read about) that can return specific information passively when hit with specific radio frequencies. Wouldn't these be more easily used than a powered device like a watch?
Anyone else know more about these technologies?
But what happens when the neighborhood/college/company bully steals your watch?
If you celebrate Xmas, befriend me (538
Sounds like the smartcards to me where you stick it in the slot & it knows your password, domain, etc. Console is locked unless you have the card.
Two roads diverged in a wood, and I - I took the one the bus load of girls just went down.
That you wear on your finger? :^)
Gimme your watch, punk!
What about using some kind of biometric data, like key cadence, or a profile of typical mouse movement characteristics (like icon overshoot?) to do it? That way its totally seamless, although one could still do some damage as it would take a few input events to establish the identity.
Sure, its not foolproof, but who wants to wear an identifying token?
http://www.masturbateforpeace.com/
What happens if you take your watch off and leave it next to the computer? It never encrypts!
Worse yet---what happens if your watch gets stolen? Now you can't get at your data! Better make sure you get the Casio watch option instead of the Breitling. No one would want to steal a Casio POS, so you should be safe.
Yeah, right.
I keep all mission-critical and government-classified information on portable USB Flash DRAM-based storage devices. They're incredibly portable and can be brought to the gym, in the car, to work, back home, swimming, hiking, biking, etc.
To be perfectly honest, I just can't bring myself to respect anyone who would leave a $4,000 laptop with supposedly top-secret information on it sitting out on a cafeteria table or something while they go sit in the bathroom and read the paper.
Just stick with portable USB drives. They're cheap, efficient, fast, and more secure than any fly-by-night research project out there right now.
If you celebrate Xmas, befriend me (538
When you stand up, hit ctrl+alt+del. When you sit down, type in your password. I had to do it at one company, and now it's just habit. Not exactly a tough thing to do. I think that these guys are trying to solve a non-problem.
The original is here. At least they waited some weeks before reposting it.
What does it actually encrypt? All sensitive data? I doubt it could do that in 5-6 seconds. Also, how do you decrypt the data if you lose your key? Or what if you fire the employee and don't get the key back? How will you get the data, then? Is there a back door for sysadmins?
Sex - Find It
A token can be easily misplaced, duplicated, or bypassed. A password is NOT a big deal to enter when you sit at your desk. If they're too lazy/clueless to enter a password, they shouldn't be responsible for any secret information.
Use a program like Scramdisk or the commercial version Drivecrypt. Keep all of your critical files on the encrypted partition. When you leave your desk, activate the screenserver with a keystroke.
Unless someone knows your password, you're safe. If they reboot, the encrypted disk is inaccessible.
What's the big deal?
While I applaud these people for making steps to make it harder to casually get information off of laptop computers, it still does not stop other attacks on such a system. Flooding the laptops area with uniformly strong signal that matches the watch's key would be as difficult as acquire-and-replicate. There seems to be a smart card like system with keys, and key encrypting keys.
It's very comprehensive, and it addresses many aspects of the social and technological attacks.
In my mind, the weak link here is clearly the watch. Watch technology isn't very complicated (read: very big), and how many designs could their possibly be? If one knows where the hardware information is located, a system replacement under the face, and you've got some issues. How many people wear watches to bed at night? Or in the shower? Difficult, but possible
A quick couple of replacements, and you have a watch that has a short range transmitter also transmitting the information that you'd need to dissolve the encryption link, and maybe begin a traditional man-in-the-middle attack. Once you see what cards the watch is holding, shouldn't the rest of the exchange be trivial?
While this is a great mechanism for an encryption scheme, what attacks are there against the physical and social component? These are the items of which spy thrillers are made, and will probably (hopefully) never come into play.
All in all, an excellent read from the UMich folk, and they have my applause.
--jaybonci
Let me use a ring, then I only lose a finger when someone wants access :~)
Guess what? I got a fever! And the only prescription.. is more cowbell!
You know, common movie elements won't understand this "token wristwatch that has a Linux-running microprocessor" thing, so let's dumb it down. How about he gets clubbed in order to get a piece of metal that has been engraved in some semi-random form so that when it's placed into its reader, it causes a door to be unlocked.
I know... call it The Key
Great, something else to buy. My fingers are cheaper and I'm not one of the people who has a problem logging in with a password. Why should I fork out cash for this?
As much as I enjoy the free publicity, this has been posted on slashdot before.
To correct a serious error that appears in this article and in the nytimes article this was cribbed from: The system was NEVER run on the IBM watch. We mentioned it as a possibility and somehow it was taken as fact.
I welcome the comments on the work, however remember that the world of university research is often more forward looking than the commercial world. That is our job!
Sounds like a nice idea. However we all know that once physical security is compromise the rest is all down hill. On-top of which, a thief that is just after the machine and cares nothing about the data will still take the machine. He doesn't know that you have a proximity sensor (whether it uses encryption or not). What I would like to see is a tool and/or system that has the kind of reliability and name recognition that something like low-jack has. What I mean is something that a crook will look at and walk away because he will recognize that it will be more trouble than it is worth. Even if he is just stealing it for the hardware. Something that he knows he just can't slap in a windows boot disk and format. Because we all know that most laptop thefts are not by criminals that want data. Its the common crook that just wants a buck. Granted what would also bring down those thefts would just be the prices in laptops coming down, the prices on those haven't fallen nearly as close to the same rate as desktops.
:)
For now I will continue to dream and maybe even write a book entitled "2085" by Ali Orwell.
I'd say why not brute force the thing, but here's something easier...Make a device that constantly scans for the signal of a token (there has to be some characteristic fingerprint to the signal). When it finds one, remember the signal and indicate to the user. User then goes and mugs target, takes laptop, uses stored signal. We've shown that man-in-the-middle attacks are do-able for a system like this, so why not keep with what works? If one knows how the system works, and can get a long enough string of interactions between the token and the server, then the key is vaunerable. Maybe this means that you have to tail the guy for a while, but let's be honest - if he's using one of these systems (I don't imagine they come cheap) then there's probably somehting worth stealing on that machine, if that's what you're up to. Make a scanner that tracks the signature of packets, walk around the financial centers of the world, and then the device goes off you know which laptops to take.
On another note, this reminds me of the plan to put RFIDs in the new high-denomination Euro-notes. Something like takes all the effort of guesing who to mug: emit the signal, and anytime you get a response, you know the guys's packing a high-value Eruo-note.
Cue The Sun...
good security should always be based on at least two of the three from the list
Something you have
Something you know
Something you are
Anything that relies on just one of these catagories is going to be significantly easier to break than one the follows the rules. Most commercial security these days is based on something you know (password) and nothing more. Good security systems require all three, biometrics, password, and a physical token. biometrics are suseptible to advanced attacks but thing like thermal imaging for skull structure combined with retinal imaging is pretty close to unbreakable. Passwords are notoriously lacking because passwords strong enough to be secure are difficult for most people to remember so they end up either weak or written down. As for token systems other than smart cards and the IBM watch I have not seen many implementations out there.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
It was discovered soon after the press release that the "zero interaction authentication" system was vulnerable to a transmission replay attack. This attack may prove fatal to the design; in any case, it should take a few years to get the kinks worked out, so don't expect it on your desktop any time soon.
I hereby place the above post in the public domain.
A possible solution is to generate a second low powered signal from the laptop; this signal would be generated from nothing more than some strongly encrypted hash, and most certainly be an AM signal. The nice thing about strong encryption is that it should be pretty much indistinguishable from random noise, so the this signal would be indistinguishable from background noise.
Then you have the frequency the signal is broadcasted on randomly shuffled based on the current time. The laptop and the token are time-synced (not a problem, most decent cryptographic tokens are time-synced anyway), so the token is always listening on the correct frequency.
At this point you have the correct waveform, although its amplitude will depend on your distance from the device. Every tenth of a second, or something, normalise the signal based on the RMS power, then compare the input signal based on what you compute it should be (you know the secret, so you can also compute the hash).
To fool this system you have to replicate the exact signal as it bounces around frequencies. Since it's bouncing around frequencies you can't just repeat the signal you're recieving on a specific frequency, since that won't matter. Further, for each part of the signal you repeat, you'll be off in intensity by a certain amount based on the frequency you're tuning into relative to the frequency its actually being transmitted at, and unless you can exactly predict the pattern you your error will vary. You can't track the frequency since you'd need to break the encryption. Really, this is nothing more than frequency scrambling that's been used by the military to secure communication for years, used in a slightly different way.
I'm sure there are other ways to solve the problem. So yes, it could be a problem if it wasn't taken into consideration, but it is a solvable problem.
As others have already mentioned, unless the article had it all wrong, it seems that you're going about this the hard way. Why not create an encrypting FS driver along the lines of Scramdisk or DriveCrypt that always stores the disk data in encrypted form and only decrypts it upon reading? The token would then simply provide the key, and when it's not present, you simply can't decrypt the data, without requiring a lengthy de/encryption process each time you leave and return? In addition, you could make the driver smart enough to let you encrypt only certain directories, plus you could still keep the cache encryption functionality as it is now.
What happens when the decryption key device fails or is lost or stolen?
I'm a netadmin for some not-very-savvy users, and if I couldn't restore access to their data just by resetting their password then they are all in trouble.
This is an issue for a lot of encryption solutions, not just this one. Is there a master key list somewhere than can be used to recover encrypted files or volumes or at least recreate the encryption key device? How long would that take? (This opens another discussion over security of the master list and key-changing and reencryption procedures for lost and stolen tokens.)
And what if the device gets stolen? I have a security token that requires a PIN in conjunction with its security (both the PIN and device are needed for access), but in the case of this article the whole point seems to be to avoid entering a password or PIN.
At the beginning of the process, the user enters a password on the watch. "That's to make sure an imposter isn't wearing your token," Noble says. Then, each second, the laptop broadcasts a cryptographic request that only the token can correctly answer. This procedure, an exchange of cryptographic numbers, is a standard security measure.
People will still use stupid passwords. GONG!. They'll use the same letter conventions that 99% of the population uses. I guarantee that one guy with a high-end laptop could walk through an office and guess 99% of the passwords within a few minutes. Or maybe they'll guess 1% and get the temp's password. Good enough, access to the internal network is almost always sufficient to own the rest of the network.
There is no technology that will override stupidity.
The fundamental problem with biometrics is that you can't change your keys. You have a set of fingerprints, retinal patterns, DNA sequences that are really pretty damn hard to change.
Biometrics can only work with strong physical security to ensure that the tests aren't being compromised (i.e., someone hacking the device).
To steal your password I have to look over your shoulder, and once done you can change it. To steal your authentication token, I have to pick your pockets, and once done you can get a new one. But I can pull your fingerprints from anything you touch, and you'll have a much, much harder time changing those.
Biometrics are often portrayed as the panacea for authentication, but of the three 'seomthing you X', it's really the weakest. Haven't we learned yet that there's no such thing as a silver bullet?
I remember reading an article about a system like this years ago - running somewhere like ARM's labs in Cambridge. They were using it for desktops rather than laptops, but that is a detail. More importantly, they had hooked a load of other systems up to the ID. It provided the security access to the building - no more fiddling for cards, the door unlocks as you approach. Rather than just blanking off the screen as you waked away from one workstation, as you moved towards another workstation, it moved your "desktop" to that station, so that your work could "follow" you round the building. And, by detecting which room you were in, the phone system could route calls to you wherever you were.
There are a lot of questions (privacy etc) about those other uses, but a system which gives you multiple returns from the single cost of wearing some kind of ID is much more likely to be adopted than a single dongle for a single job.
Consciousness is an illusion caused by an excess of self consciousness.