Slashdot Mirror
Backdoor Found In China-Made US Military Chip?
Hugh Pickens writes "Information Age reports that the Cambridge University researchers have discovered that a microprocessor used by the US military but made in China contains secret remote access capability, a secret 'backdoor' that means it can be shut off or reprogrammed without the user knowing. The 'bug' is in the actual chip itself, rather than the firmware installed on the devices that use it. This means there is no way to fix it than to replace the chip altogether. 'The discovery of a backdoor in a military grade chip raises some serious questions about hardware assurance in the semiconductor industry,' writes Cambridge University researcher Sergei Skorobogatov. 'It also raises some searching questions about the integrity of manufacturers making claims about [the] security of their products without independent testing.' The unnamed chip, which the researchers claim is widely used in military and industrial applications, is 'wide open to intellectual property theft, fraud and reverse engineering of the design to allow the introduction of a backdoor or Trojan', Does this mean that the Chinese have control of our military information infrastructure asks Rupert Goodwins? 'No: it means that one particular chip has an undocumented feature. An unfortunate feature, to be sure, to find in a secure system — but secret ways in have been built into security systems for as long as such systems have existed.'" Even though this story has been blowing-up on Twitter, there are a few caveats. The backdoor doesn't seem to have been confirmed by anyone else, Skorobogatov is a little short on details, and he is trying to sell the scanning technology used to uncover the vulnerability.
It sells...
Even if this case turns out to be a false alarm, allowing a nation that you repeatedly refer to as a 'near-peer competitor' to build parts of your high-tech weaponry is idiotic.
The original article is here.
It refers to an Actel ProAsic3 chip, which is an FPGA with internal EEPROM to store the configuration.
The determined Real Programmer can write Fortran programs in any language.
Either the claims will be backed up by independently reproduced tests or they won't. But, given his apparent track record in this area and the obvious scrutiny this would bring, Skorobogatov must have been sure of his results before announcing this.
Here's his publications list from his University home page, FWIW:
http://www.cl.cam.ac.uk/~sps32/#Publications
Burns: We're building a casino!
McAllister: Arrr. Give me 5 minutes.
From TFA:
Today we released the drafts of our full papers on QVL technology due to accidental publicity, because someone put the link to our very old drafts of abstracts on Reddit.
This is a security guy I would trust, yessir.
Not sure how exciting this is, as they needed physical access to the chip to get anything out of it.
That entire article reads more like a press release with FUD than anything with any facts.
Which chip?
Which manufacturer?
Which US customer?
No facts and LOTS of claims. It's pure FUD.
(Not that this might not be a real concern. But the first step is getting past the FUD and marketing materials and getting to the real facts.)
Chinese leaders are in a cold war with the west. As such, it is far cheaper and easier to be able to shut down an adversaries equipment if you are manufacturing it for them. If the west would quit being foolish, they would insist on equipment made in secured companies. And Google has already proved that nothing in China is secured from the gov.
I prefer the "u" in honour as it seems to be missing these days.
The back-door described in the white paper requires access to the JTAG (1149.1) interface to exploit. Most deployed systems do not provide an active external interface for JTAG. With physical access to a "secure" system based upon these parts, the techniques described in the white paper allow for a total compromise of all IP within. Without physical access, very little can be done to compromise systems based upon these parts.
Sun Tzu said the greatest victory is one which doesn't require a shot. One won by subverting the enemy from within.
What greater subversion can there be than to convince the enemy to hire you to build their weapon's systems components?
Apparently the American Military (and probably that of the rest of the world) hasn't bothered reading any "classic" literature on warfare before signing on the dotted line...
I do not fail; I succeed at finding out what does not work.
1) Read the paper http://www.cl.cam.ac.uk/~sps32/Silicon_scan_draft.pdf
2) This is talking about FPGAs designed by Microsemi/Actel.
3) The article focuses on the ProAsic3 chips but says all the Microsemi/Actel chips tested had the same backdoor including but not limited to Igloo, Fusion and Smartfusion.
4) FPGAs give JTAG access to their internals for programming and debugging but many of the access methods are proprietary and undocumented. (security through obscurity)
5) Most FPGAs have features that attempt to prevent reverse engineering by disabling the ability to read out critical stuff.
6) These chips have a secret passphrase (security through obscurity again) that allows you to read out the stuff that was supposed to be protected.
7) These researchers came up with a new way of analyzing the chip (pipeline emission analysis) to discover the secret passphrase. More conventional anaylsis (differential power analysis) was not sensitive enough to reveal it.
This sounds a lot (speculation on my part) like a deliberate backdoor put in for debug purposes, security through obscurity at it's best. It doesn't sound like something secret added by the chip fab company, although time will tell. Just as embedded controller companies have gotten into trouble putting hidden logins into their code thinking they're making the right tradeoff between convenience and security, this hardware company seems to have done the same.
Someone forgot to tell the marketing droids though and they made up a bunch of stuff about how the h/w was super secure.
Absolutely. The US military should have a strict policy of only buying military parts from sovereign, free, democratic countries with a long history of friendship, such as Israel, Canada, Europe, Japan and South Korea.
And a preference should be given to American-made parts, since you need domestic factories to mobilise in times of war.
First problem..... they already have that policy. But the problem is that the components used for military and government applications have to be purchased from American companies. Then to save a buck, the companies sub-contract for components from places like China and "assemble" the equipment in friendly countries. That way, the product does not have a "made in China" sticker on them.
Second problem.... 20 years ago the DOD had their own processor manufacturing facilities, IC chips, etc. They were shut down in favor of commercial equipment because some idiot decided it was better to have an easier time buying replacement parts at Radioshack than buying quality military-grade components that could last in austere environments. (Yes, speaking from experience). Servers and workstations used to be built from the ground up at places like Tobyhanna Army Depot. Now, servers and workstations are bought from Dell.
sudo make me a sandwich
Have gnu, will travel.
Fabs are expensive. The latest generation nodes cost billions of dollars to set up and billions more to run. If they aren't cranking chips out 24/7, they're literally costing money. Yes, I know it's hte military, but I'm sure people have a hard time justifying $10B every few years just to fab a few chips. One of the biggest developments in the 90s was the development of foundries that let anyone with a few tens of millions get in the game of producing chips rather than requiring billions in startup costs. Hence the startup of tons of fabless companies selling chips.
OK, another option is to buy a cheap obsolete fab and make chips that way - much cheaper to run, but we're also talking maybe 10+ year old technology, at which point the chips are going to be slower and take more power.
Also, building your own computer from the ground up is expensive - either you buy the designs of your servers from say, Intel, or design your own. If you buy it, it'll be expensive and probably require your fab to be upgraded (or you get stuck with an old design - e.g., Pentium (the original) - which Intel bought back from the DoD because the DoD had been debugging it over the decade). If you went with the older cheaper fab, the design has to be modified to support that technology (you cannot just take a design and run with it - you have to adapt your chip to the foundry you use).
If you roll your own, that becomes a support nightmare because now no one knows the system.
And on the taxpayer side - I'm sure everyone will question why youre spending billions running a fab that's only used at 10% capacity - unless you want the DoD getting into the foundry business with its own issues.
Or, why is the military spending so much money designing and running its own computer architecture and support services when they could buy much cheaper machines from Dell and run Linux on them?
Hell, even if the DoD had budget for that, some bean counter will probalby do the same so they can save money from one side and use it to buy more fighter jets or something.
30+ years ago, defense spending on electronics formed a huge part of the overall electronics spending. These days, defense spending is but a small fraction - it's far more lucrative to go after the consumer market than the military - they just don't have the economic clout they once had. End result is the miliary is forced to buy COTS ICs, or face stuff like a $0.50 chip costing easily $50 or more for same just because the military is a bit-player for semiconductors.
This is a physical-access backdoor. You have to have your hands on the hardware to be able to use JTAG. It's not a "remote kill switch" driven by a magic data trigger, it's a mechanism that requires use of a special connector on the circuit board to connect to a dedicated JTAG port that is simply neither used nor accessible in anything resembling normal operation.
That said, it's still pretty bad, because hardware does occasionally end up in the hands of unfriendlies (e.g., crashed drones). FPGAs like these are often used to run classified software radio algorithms with anti-jam and anti-interception goals, or to run classified cryptographic algorithms. If those algorithms can be extracted from otherwise-dead and disassembled equipment, that would be bad--the manufacturer's claim that the FPGA bitstream can't be extracted might be part of the system's security certification assumptions. If that claim is false, and no other counter-measures are place, that could be pretty bad.
Surreptitiously modifying a system in place through the JTAG port is possible, but less of a threat: the adversary would have to get access to the system and then return it without anyone noticing. Also, a backdoor inserted that way would have to co-exist peacefully with all the other functions of the FPGA, a significant challenge both from an intellectual standpoint and from a size/timing standpoint--the FPGA may just not have enough spare capacity or spare cycles. They tend to be packed pretty full, 'coz they're expensive and you want to use all the capacity you have available to do clever stuff.
FPGAs commonly protect user-code with encryption. An encryption engine is included in the silicon to which the user has limited access to crypto=keys with which to encrypt the code that is installed in ROM/Flash.
A number of attacks are known against microcontrollers/FPGAs that secure code with encryption - notably differential power analysis (DPA) which works by connecting a current probe to the chip, and collecting measurememnts of energy consumption as the device performs an authentication operation. By carefully, measuring power traces over thousands of authentication operations, statistical analysis can reveal clues about the internal secret keys; potentially allowing recovery of the key within useful periods of times (minutes to hours).
These secure FPGAs contain a heavily obfuscated hardware crypto-engine, with lots of techniques to obstruct DPA (deliberately unstable clocks, heavy on-chip RC power filtering, random delay stages in the pipeline, multiple "dummy" circuits so that an operation which would normally require fewer transistors than an alternative, has its transistor count increased, etc.). The idea being that these countermeasures reduce the DPA signal and increase the amount of noise, making recovery of useful statistics impractical. In their papers, this group admit that the PA3 FPGAs are completely impervious to DPA, with no statistical clues obtained even after weeks of testing.
This group have developed a new technique which they call PEA which is a much more sensitive technique. It involves extracting the FPGA die, and mapping the circuits on it - e.g. using high-resolution infra-red thermography during device operation to identify "interesting" parts of the die by heat production under certain tasks - e.g. caches, crypto pipelines, etc. Having identified interesting areas of the die, an infra-red microscope with photon counter is focused on the relevant circuit area. As it happens, transistors glow when switched, emitting approx 0.001 photons per switching operation. The signal from the photon counter is therefore analogous to the DPA signal, but with a much, much stronger signal-to-noise ratio, allowing statistical analysis with far fewer tries. The group claim the ability to extract the keys from such a secure FPGA in a few minutes of probing with authentication requests.
The researchers claim to have found the backdoor, by fuzzing the debug/programming interface, and finding an undocumented command that appeared to trigger a cryptographic authentication. By using their PEA technique against this command, they were able to extract the authentication key, and were able to open the backdoor, finding they were able to directly manipulate protected parameters of the chip.
http://erratasec.blogspot.com/2012/05/bogus-story-no-chinese-backdoor-in.html
Bogus story: no Chinese backdoor in military chip
"Today's big news is that researchers have found proof of Chinese manufacturers putting backdoors in American chips that the military uses. This is false. While they did find a backdoor in a popular FPGA chip, there is no evidence the Chinese put it there, or even that it was intentionally malicious.
Furthermore, the Actel ProAsic3 FPGA chip isn't fabricated in China at all !!
Muchas Gracias, Señor Edward Snowden !