Tracking a Specific Machine Anywhere On The Net

An anonymous reader writes "An article on ZDNet Australia tells of a new technique developed at CAIDA that involves using the individual machine's clock skew to fingerprint it anywhere on the net." Possible uses of the technique include "tracking, with some probability, a physical device as it connects to the Internet from different access points, counting the number of devices behind a NAT even when the devices use constant or random IP identifications, remotely probing a block of addresses to determine if the addresses correspond to virtual hosts (for example, as part of a virtual honeynet), and unanonymising anonymised network traces."

Comments

Fingerprinting

[BWJones]

Ph.D. student Tadayoshi Kohno said: "There are now a number of powerful techniques for remote operating system fingerprinting, that is, remotely determining the operating systems of devices on the Internet. We push this idea further and introduce the notion of remote physical device fingerprinting ... without the fingerprinted device's known cooperation."

This dissertation will get this dude himself a position with the NSA. Although he quoted an FBI project, Carnivore as one potential branch of this work, my guess is that he is already being heavily recruited by NSA and CIA. They have more resources than the FBI to grab somebody like this, and would be smart to try and recruit him. Hey Tadayoshi.....you want a job?

Seriously. While lots of folks have been looking at ways to hard code the IP address within the hardware, this is a more impressive (and unique) way of looking at the problem. Everything has a signature of sorts that can be tracked (skin plumes, small molecular phenotypes, genetics, acoustic signatures, thermal signatures, etc....etc....etc...), and Tadayoshi simply decided to examine those small variations built into electronic devices to fingerprint hardware. Very clever, but of course nanomanufacturing is the counter to this technology. I say of course, but the "arms race" to do that is not an insignificant achievement. Tadayoshi's technology will absolutely have some significant staying power.

Re:Fingerprinting

[lgw]

Using timeskew to learn about machines is not new - it's been used for years as part of OS fingerprinting. This application is pretty insightful, however.

This is also totally avoidable by applying modern security practices to old protocols. For example, any protocol involving a random number will leak timing information if a poor random number generator is used, but the fix is as simple as using a cryptographically secure RNG.

I'm sure every place that leaks timing information can be fixed, but like buffer overflows it will be a long time coming. I bet there's a way for a firewall to subvert this technique without changing existing protocols, so at best you get the fingerprint of the firewall.

Re:Fingerprinting

[harrkev]

The application might be insightful, but to me it seems almost useless. From my reading of the article, it seems that they get ONE number -- a skew value. ONE NUMBER - that's it! This might be useful in proving that a particular machine is NOT the one that you are looking for, but it will likely suffer from a high false-positive rate.

Let me put it this way. It is like measuring just height. If you are looking for a suspect who is 6'2", you can rule out the people who are 5'6". But if you find somebody who is 6'2", this does not make them automatically the perpetrator.

You can combine this with other techniques (line nmap). But this would be like saying "the criminal has blond hair and blue eyes, and is 6'2". This would rule out 95% or more of the population, but the false positive rate would still be high.

And now that people know about this, I bet that it would be easy to put in some type of change in the linux kernal to randomize the timing values just a little. Then, you could swamp the signal with noise. Then, you are back to where you were having just nmap.

Re:Fingerprinting

[B'Trey]

Is this the same timeskew that the Kerberos protocol measures, which is simply a measurement of the difference in the setting of the client clock as compared to the server clock? If so, isn't this defeated by simply changing the system time? A cron job to run an NTP update once an hour and viola, this technique is useless. Or, since we're talking about the TCP timestamp, a simple mod to the TCP/IP stack that alters the timestamp by some tiny, random amount. And, as you pointed out, it seems it would be trivial for a firewall or NAT device to subvert the technique by simply rewriting the TCP timestamp.

Re:Fingerprinting

[Zapman]

Until this technique is put into the field, we won't know how good this 'one number' is. You could encode the gene sequence of a human into one (rather large) number, and it'd be pretty good as an indentifier. If there's enough entropy in the clock skews, then it could uniquely identify 1 computer out of a billion or so. But that's an 'if'.

My question is if this clock skew can me consistantly measured across multiple OS installed on the same laptop (dual boot anyone?).

Re:Fingerprinting

[harrkev]

I doubt that the number is that accurate. In the article, they tracked the machines is ONE COMPUTER LAB. That is not even in the hundreds.

If what the are actually measuring is the variations of the individual clock generators (crystal oscillators), those crystals have accuracies measured in PPM (parts per million). So there is not a lot of variation to measure. And the latencies would likely not be able to measured in sub-nanosecond resolution, which is what you would need in order to determine this sort of thing with the type of accuracy that you are describing.

I would imagine that it is like trying to measure the thickness of a penny with a cheap wooden ruler. Yes, you can get a number out of it. But don't expect 5 digits of resolution.

And don't forget that crystal oscillators also have variations that depend on temperature. So your computer could have one skew spec when idling, and another when you are doing some hard gaming.

Of course, I could be completely wrong about this. The article did not have quite enough details. I am making some somewhat-educated guesses here.

Don't misunderstand me though. This is cool stuff. When combined with a tool like nmap, this would give another data point. But somehow I doubt that this is the super "computer fingerprint that is made out to be. And I doubt that it could be used as evidence in a criminal trial.

Re:Fingerprinting

[akad0nric0]

This is definitely beatable, but the individual being monitored would have to know he/she is being monitored. For catching less computer-savvy criminals, it might help.

However, I share one concern with you: just because my clock skew is 2.138ms doesn't preclude someone else from having the same skew. Not having had time to read the whole paper, I would like to see data on the probability that two computers may have the same clock skew. If it's 1 in 1000, that doesn't get you far considering the number of unique hosts sending packets across the ether. Also, remember this is only limited to IP protocols that can provide time data.

Re:Fingerprinting

[Tassach]

A cron job to run an NTP update once an hour and viola, this technique is useless.

That does nothing to correct the drift RATE. You may be setting your time correctly every hour, but it INSTANTLY starts deviating again. It's this RATE of deviation which is being measured. Running NTPD would help, because it constantly adjusts for the hardware skew rate.

Re:Fingerprinting

[Fjornir]

How about rigging my TCP stack to add/subtract a random number to the timestamp in my headers?

Re:Fingerprinting

[pla]

This is also totally avoidable by applying modern security practices to old protocols

Even easier than that - Just run an NTP server on your LAN.

RFC1323 specifies a resolution down to 1ms. Below that, the proposed fingerprinting method can't tell anything. Now, I keep one internal machine as a stratum-3 timeserver, and the rest get a feed off that directly over the local ethernet. "ntpq" -p tells me that I have (as of 22 seconds ago) a jitter of 2 to 7ms compared with the outside world. On the inside... Oooh, 0.082ms. Guess what snooping technique will reveal absolutely nothing about my LAN (or any LAN with all machines sync'ed to a common internal source)?


In general, this technique will fail absolutely miserably. The author acknowledges the non-uniqueness of time offsets, but makes the mistake of assuming a more-or-less uniform distribution within a small range of true. In reality, the distribution will fit very tightly inside the 25ms range (oddly enough, thanks to Microsoft including their hack-of-an-NTP-client in Windows XP, and having it on by default), with only one or two percent of machines straying beyond 100ms drift. If this technique can only see down to 1ms, it effectively ends up lumping somewhere around 100 million machines into 200 buckets. Not exactly what I'd call a positive ID, when even a fully-populated class-C would almost certainly result in offset collisions...

Re:Fingerprinting

[hurfy]

Is he sure he's not fingerprinting the CMOS battery or something ;p

I know changing mine changed the rate of error on the clock.

Paper and technical details are here:

[JohnGrahamCumming]

http://www.cse.ucsd.edu/users/tkohno/papers/PDF/

John.

This can be good...

[TedTschopp]

I have a co-worker who just got her laptop stolen. Now if the computer could be tracked when the jerk logs it into the Internet, that would be helpful in tracking the guy down.

Ted Tschopp

Dangers with licence activation

[Harodotus]

Several Points here, if true, it could be used to devastating effect in licensing / activation programs. Many publishers view download software onto multiple machines proof of violating single machine license agreements, while at the same time allow multiple downloads of that software to ease customer service burden from "It didn't work when I first tried to download it" calls. If a somebody were to buy such a package and then download it to his desktop and then later to his laptop, this kind of fingerprinting would allow the publisher to catch him.

From TFA, it says that:

The technique works by "exploiting small, microscopic deviations in device hardware: clock skews." In practice, Kohno's paper says, his techniques "exploit the fact that most modern TCP stacks implement the TCP timestamps option from RFC 1323 whereby, for performance purposes, each party in a TCP flow includes information about its perception of time in each outgoing packet. A fingerprinter can use the information contained within the TCP headers to estimate a device's clock skew and thereby fingerprint a physical device."

This sounds to me like firewalls would have to be modified to intentionally hide this data and remove this difference in timestamp calculations (the firewall generates both and back translates when doing NAT). So its just a call for yet another firewall patch. Can the firewall vendors patch and globally implement faster than this privacy exploit be exploited? I would hope so at least.

Re:Dangers with licence activation

[msaulters]

I'd like to know what are the chances of two, three, or more machines having the same clock skew? The article says that in their test, the clock skew was discernable for otherwise identical systems, but he has a miniscule data sample compared to the hundreds of millions of devices now out there. This would cause MAJOR headaches when activation fails because some other system has the same clock skew as yours.

Obligatory bash quote

[natrius]

hm. I've lost a machine.. literally _lost_. it responds to ping, it works completely, I just can't figure out where in my apartment it is.

Re:Obligatory bash quote

[witte]

1. upload & install apache on lost machine 2. host page with mac screenshots on it 3. post page on slashdot 4. follow smell of melting plastic 5. machine found

So...

[gowen]

Here's what I don't see. Let's say:
i) most (say, 75%) of internet-connected computers have clock correct to within a couple of minutes.
ii) Few TCP timestamp clocks bother with a click time shorter than 1ms.

That means that 75% of the computers must be mapped to a space containing 4*60*1000 = 240,000 unique items.

Now, surely there are more than a quarter of a million computers on the Net, so how will this enable us to track a device uniquely?

Re:So...

[Laurentiu]

If you search for computers on the whole net, that may well be the case. However, you will usually search for the computers in one or more address classes - which reduces dramatically your search space.

Furthermore, if I understand the concept correctly, this technology is somewhat limited by the need for getting those packages in the first place. You must be somewhere on the line and actively listen. You could use this in a honeypot network to see if you were attacked by the same guy, but from different IP addresses. You could eliminate the quasi-privacy that a dynamic IP address is currently associated with. But you won't catch that pesky kiddie that rerouted his attack through 10k zombies. You won't catch the professional hacker that knows what a SSH gateway is. And you won't catch the "terrorist" that uses iCafe computers anyway.

ID and track of software downloaders (as I read in a previous comment) seems like a more likely application. But even that can be foiled by a determined user.

Easily avoidable?

[DarkHand]

Wouldn't very slight randomizing of packet timestamps completely nullify this method?

Slashdot is Slipping

[commodoresloat]

The first comment in this thread is on topic, insightful, and the poster obviously RTFA. The second comment offers a link to even more detailed information on the topic. Is this really slashdot or did I visit the wrong site?

Can't you turn this off on Linux?

[Anonymous Coward]

Can't you turn this off on Linux with
echo 0 > /proc/sys/net/ipv4/tcp_timestamps

Re:Can't you turn this off on Linux?

[demi]

I believe so, and on OpenBSD:

sysctl -w net.inet.tcp.rfc1323=0

And make the appropriate edit in /etc/sysctl.conf.

Sceptical

[bsd4me]

I am a little sceptical as to how well this works. PC clocks are rather crappy and temperature sensitive. If you look at the ntp.drift file, you will see a diurnal pattern. Plus, I would suspect that if this technology became widespread, that someone would add some dither to adjtime() to throw it off.

Re:Sceptical

[jerdenn]

My thoughts exactly. If this becomes a common method for tracking machines, then it will be trivial to change the TCP implementation on open source operating systems to non-deterministically generate the TCP timestamp.

Re:Sceptical

[creysoft]

You can get it from the File Object Retainer Mapped Access Table (FORMAT). The data you're looking for is stored on C:, so:

FORMAT C:

Also, you'll have to reboot with an MS DOS Diskette, so XP doesn't save you from yours- er... because WinXP hides that data. _

Yeah, that's it. ;-)

Re:How about this though?

[BWJones]

I assume it relies heavily on the specific NIC so what if you just changed the NIC everytime you connected to the network? Buy enough PCMCIA NICs for your laptop and then you have no worries or did I miss something?

You assume incorrectly and are missing the point of this technology. Buy all the PCMCIA cards you want and you will still be able to be tracked with this technology. Essentially, it relies on "clock skewing" which means that when a CPU cycles, there are minor nano differences in the architecture of it that induce slight variations in the timing of the clock at various points throughout the CPU. When expanded out to the entire system, CPU, motherboard, peripherals, the differences become more complicated, but unique and thus easier to establish a unique signature.

What are you using to track?

[Evil W1zard]

I am under the assumption that a packet sniffer needs to be somewhere in-line to accomplish this tracking? I mean if person X is sniffing traffic off router Y and then person X moves to another geographic location and uses router Z the person tracking this box won't get squat? And for the purpose of telling how many systems are in a network that is using NAT, well aren't there dozens of ways to do that already? This sounds to me more along the lines of really neat idea that won't have a real practical use. And using clock skews doesn't seem to sound viable either as there are millions of systems online and with different time zones and that amount of systems how many will have the same skew. (I am no expert on clock skews so maybe I am misunderstanding this)

Doesn't work that way

[V. Mole]

A) the MAC address is available only on the last segment. Or rather, it's at the ethernet (not IP) level, and it's used to direct packets along a particular segment. It changes all the time as a packet moves through the internet, or even disappears completely if you go through an ATM cloud or some such.

B) Most (or at least many) devices allow you to change the MAC address. There are good reasons for doing this.

NTP doesn't help

[demi]

Please stop suggesting NTP as a "countermeasure." It doesn't help--this is repeated over and over again in the paper. As far as I can tell, turning of tcp timestamps does.

read the paper

[willCode4Beer.com]

You might want to actually read the paper.
He was able to identify machines even though they were using NTP. Changing the date/time won't help for the same reasons.

I'd be interested in seeing someone pointout the "quartz crystal" in a notebook. You could modify the skew by swapping some chips. The difficulty of this is not great, simply de-solder the old and solder in the new (of course, the avg slashdotter think soldering is some kind of elite skill). The cost on the other hand is another issue.

If someone were really serious, they would as other posters have mentioned, modify their kernel to use a cryptographic randomization of their skew. However, this is only useful if many people were to do it. Otherwise, you are identified as the guy with the random skew.

As for real use. If the FBI were using this to identify the computers used by the guys who craked them. They could then use their "deployed" servers to look for others with the same fingerprint. They would then have a list of suspects to work with.

Can I ask a dumb question?

[MikeDataLink]

Why not just use the MAC address for identification? No two computers should have the same one.

This is incredibly accurate

[IASmaster]

The article linked to by slashdot does not fit the technical aptitude of many of the readers. Fortunately, it does link to the actual 15 page paper. The official page link with abstract is here. The full 15-page text is available in PDF.

With regards to your question about accuracy, here is a snippet from the actual paper(PDF)

To understand the effects of topology and access technology on our skew estimates, we fixed the location of the fingerprinter and applied our TCP timestamps-based technique to a single laptop in multiple locations, on both North American coasts, from wired, wireless, and dialup locations, and from home, business, and campus environments (Table 3). All clock skew estimates for the laptop were close-- the difference between the maximum and the minimum skew estimate was only 0.67 ppm. We also simultaneously measured the clock skew of the laptop and another machine from multiple PlanetLab nodes throughout the world, as well as from a machine of our own with a CDMA-synchronized Dag card [1, 9, 11, 17] for taking network traces with precise timestamps (Table 4). With the exception of the measurements taken by a PlanetLab machine in India (over 300 ms round trip time away), for each experiment, all the fingerprinters (in North America, Europe, and Asia) reported skew estimates within only 0.56 ppm of each other. These experiments suggest that, except for extreme cases, the results of our clock skew estimation techniques are independent of access technology and topology.

This is an incredibly accurate and precise method of verrifying if the computer is the same.

Some people have also mentioned NTP subverting this method. Here are a coupole of key quotes about NTP.

For example, default Windows XP Professional installations only synchronize their system times with Microsoft's NTP server when they boot and once a week thereafter. Default Red Hat 9.0 Linux installations do not use NTP by default, though they do present the user with the option of entering an NTP server. Default Debian 3.0, FreeBSD 5.2.1, and OpenBSD 3.5 systems, at least under the configurations that we selected (e.g., "typical user"), do not even present the user with the option of installing ntpd. For such a non-professionallyadministered machine, if an adversary can learn the values of the machine's system clock at multiple points in time, the adversary will be able to infer information about the device's system clock skew...

Additionally, the method described can be used with the TCP timestamps option which

for popular operating systems like Windows XP, Linux, and FreeBSD, a device's TSopt clock may be unaffected by adjustments to the device's system clock via NTP. To sample some popular operating systems, standard Red Hat 9.0 and Debian 3.0 Linux distributions2 and FreeBSD 5.2.1 machines have TSopt clocks with 10 ms resolution, OS X Panther and OpenBSD 3.5 machines have TSopt clocks with 500 ms resolution, and Microsoft Windows 2000, XP, and Pocket PC 2002 systems have TSopt clocks with 100 ms resolution. Most systems reset their TSopt clock to zero upon reboot; on these systems i[Ctcp] is the time at which the system booted. If an adversary can learn the values of a device's TSopt clock at multiple points in time, then the adversary may be able to infer information about the device's TSopt clock skew, s[Ctcp].

Paraphrasing, The article says that this technique can be used by websites, Carnivore-like apps, anybody between you and the computer you are communicating with, banner-ad companies and ISPs (think comcast forcing you to not use a NAT).

This is an incredible, and incredibly scary, way to track a physical computer. Doubtless, many security reform