Chinese manufacturer puts hardware backdoor onto Supermicro server boards.

[floobydust]

I find it believable for the PLA military to add a little IC for a backdoor.

Intel Management Engine is in all their server chipsets, used on Supermicro blades. The PCB artwork was modified and a part "added" according to the press out there.
Example, for the C600 series server chipsets, the ME F/W is loaded from SPI flash. Six-pins (EMI filter package ) are enough for an MCU with SPI to communicate with the chipset or main CPU's and inject a Trojan.
Or maybe the chinese simply added an SPI flash IC.

[bd139]

But again two questions;

1. Why? There are much easier attack vectors.
2. Who? Supermicro stuff is designed in USA. Do they not do design validation on production runs and sampling?

[mtdoc]

But what Bloomberg describes is, as you said, unicorns. Forcing the OS to do what? Which OS after all? BIOS, EFI and all that crap is not running when a proper OS is in execution.

Then read this  https://www.zdnet.com/article/minix-intels-hidden-in-chip-operating-system/

MINIX also has access to your passwords. It can also reimage your computer's firmware even if it's powered off. Let me repeat that. If your computer is "off" but still plugged in, MINIX can still potentially change your computer's fundamental settings.

But doesn't this support the idea that there are easier ways to do this than add a new chip to a motherboard?  Wouldn't altering or cloning/replacing an existing chip be easier and require fewer potential points of detection in the supply chain?

I don't doubt the Chinese are motivated and capable to do such spying (as are Western intelligence agencies) but the nature and timing of this report seems awfully suspicious.  Another possibility is that this has been known for sometime but the report has been sat on, awaiting the proper geopolitical timing.

[wraper]

But again two questions;

1. Why? There are much easier attack vectors.
2. Who? Supermicro stuff is designed in USA. Do they not do design validation on production runs and sampling?

I dunno if it's real or not but there are good reasons to do it. They cannot plant hacked firmware on a few devices for particular customer. They cannot intercept a few packages for particular customer and do it NSA style. If they do this in wide scale hoping it gets somewhere where they need, hacked firmware won't do. Too wide exposure and someone will find it. With such approach it needs to be something extremely difficult to find.

[floobydust]

"The majority of its workforce in San Jose is Taiwanese or Chinese"
The PCB assembler, was supplied modified PCB's and stuffed on the extra one little tiny part.

I can't think of a better or easier way to surreptitiously get a trojan into servers across the world.

[bd139]

But again two questions;

1. Why? There are much easier attack vectors.
2. Who? Supermicro stuff is designed in USA. Do they not do design validation on production runs and sampling?

I dunno if it's real or not but there are good reasons to do it. They cannot plant hacked firmware on a few devices for particular customer. They cannot intercept a few packages for particular customer and do it NSA style. If they do this in wide scale hoping it gets somewhere where they need, hacked firmware won't do. Too wide exposure and someone will find it. With such approach it needs to be something extremely difficult to find.

That’s why software is the place to do it.

I suggest people read the following:

1. https://www.archive.ece.cmu.edu/~ganger/712.fall02/papers/p761-thompson.pdf
2. https://www.theregister.co.uk/2010/12/15/openbsd_backdoor_claim/

Successful attacks leave no evidence and are plausibly deniable. Hardware is a massive steaming chunk of curly evidence.

"The majority of its workforce in San Jose is Taiwanese or Chinese"
The PCB assembler, was supplied modified PCB's and stuffed on the extra one little tiny part.

I can't think of a better or easier way to surreptitiously get a trojan into servers across the world.

Software.

[Halcyon]

But again two questions;

1. Why? There are much easier attack vectors.
2. Who? Supermicro stuff is designed in USA. Do they not do design validation on production runs and sampling?

I dunno if it's real or not but there are good reasons to do it. They cannot plant hacked firmware on a few devices for particular customer. They cannot intercept a few packages for particular customer and do it NSA style. If they do this in wide scale hoping it gets somewhere where they need, hacked firmware won't do. Too wide exposure and someone will find it. With such approach it needs to be something extremely difficult to find.

That’s why software is the place to do it.

I suggest people read the following:

1. https://www.archive.ece.cmu.edu/~ganger/712.fall02/papers/p761-thompson.pdf
2. https://www.theregister.co.uk/2010/12/15/openbsd_backdoor_claim/

Successful attacks leave no evidence and are plausibly deniable. Hardware is a massive steaming chunk of curly evidence.

"The majority of its workforce in San Jose is Taiwanese or Chinese"
The PCB assembler, was supplied modified PCB's and stuffed on the extra one little tiny part.

I can't think of a better or easier way to surreptitiously get a trojan into servers across the world.

Software.

+1 software, absolutely.

[wraper]

Software.

You can relatively easily find that software/firmware was altered. If it's a tiny innocent looking EMI filter, sleeping until comes time to do it's dirty job, it's way more difficult to figure it out.

[bd139]

Can you tell a malicious alteration from non malicious code?

Hell no. We’ve had auditors walk straight over stuff we threw in to trip them up and we’re not experts in that sort of thing (well not intentionally  )

There’s no formal verification of the code. If they can get a massive amount of people in a supply chain then they have the funds and resources to get a software plant or take over a company network. And you can do it without leaving a speck of evidence. An actual tangible device is a smoking gun of immense scale and a point of analysis.

Look at the recent BA MageCart hack as an example.

There is no logic.

[wraper]

Can you tell a malicious alteration from non malicious code?

Hell no. We’ve had auditors walk straight over stuff we threw in to trip them up and we’re not experts in that sort of thing (well not intentionally  )

You can simply start verifying if data matches to what is supposed to be there. If something is off, it's a signal for further investigation.

[Halcyon]

Can you tell a malicious alteration from non malicious code?

Hell no. We’ve had auditors walk straight over stuff we threw in to trip them up and we’re not experts in that sort of thing

Hash values would be a quick and easy start.

If your auditors can't manage a basic thing like that, maybe you should reconsider who audits your gear.

[Rick Law]

china tries to recruit spies on linkedin. enough said

Everyone spies on everyone else.  Sometimes we forget, developed nation spy on each other too.  Remember when France spies was bugging Air France first-class for business/industrial info?^{See [1]}  So, I would not be too surprised by or being too "holier than thou" with China being in the spy game.

That said, perhaps we may be too preoccupied with spying.  I wonder could it be some copy-protection/detection like some of the useless bits of circuit put into an IC just to detect if this is fake manufactured by others.


(For those who missed the news back then...)
[1] "Les unsporting gits! French spies BUGGED Concorde passengers", The Register, 2015

"Ed Wallace, director of incident response and advanced threats at security consultancy MWR Infosecurity, pointed us towards forgotten reports that French spies routinely bugged first-class passengers flying with Air France – including Concorde passengers – back in the 1990s.

The idea was that business people relaxing on a long trip, and perhaps enjoying a drink or two, might discuss all manner of commercially sensitive information, oblivious to the possibility of eavesdropping. The motive was economic espionage rather than national security or anti-terrorism."

Quoted from:
https://www.theregister.co.uk/2015/06/09/french_spied_concorde_passengers/

[bd139]

Can you tell a malicious alteration from non malicious code?

Hell no. We’ve had auditors walk straight over stuff we threw in to trip them up and we’re not experts in that sort of thing (well not intentionally  )

You can simply start verifying if data matches to what is supposed to be there. If something is off, it's a signal for further investigation.

That is formal verification. You can’t retrospectively use this to apply security. You have to build it in from the start. You can’t start with a sieve and fill all the holes with cement.

Can you tell a malicious alteration from non malicious code?

Hell no. We’ve had auditors walk straight over stuff we threw in to trip them up and we’re not experts in that sort of thing

Hash values would be a quick and easy start.

If your auditors can't manage a basic thing like that, maybe you should reconsider who audits your gear.

Hashing is performed on the build output. Hashing also requires the hashes to be stored. The attack vectors are still valid at source level. And bypassing hash verification and even changing the hashes are viable target attacks. In fact they are some of the easiest. Look up “goto fail” for a fine example of a suitably obscure cock up that stopped validation.

Multiple auditors for reference. It was not identified because it’s is subtle, requires extremely large knowledge of the problem domain, and a specific code path to be executed. An external auditor or internal auditor would need to be a domain expert with several years experience to do it.

We joke that When your mid level dev turns up with a model s Tesla then you need to read all his commits in the last year again.

This is my bread and butter. I get paid to deal with this. You need to come up with some better responses. Sorry. People really do not understand secure programming models.

[TimNJ]

So, if Supermicro is an American company and designed the board in the US (and/or the US team has acceaa to the manufacturing files), shouldn’t they be able to confirm that chip U1xx was not part of the design but appears on the manufactured product?

[mtdoc]

Some interesting, informed discussion of the technical aspects of this story in this Twitter thread.

[ogden]

That article reeks of bullshit and FUD to me.
Unfortunately written by someone who doesn't understand the tech, losing any details that might be informative.
Something inline with SPI flash is about the only thing I can guess based on the sparse info there, maybe even just disabling any write protection.

Though one detail was mentioned: "The illicit chips could do all this because they were connected to the baseboard management controller".  I find it plausible that they attacked WPCM450 and did it by putting their spy chip on SMBus.

[Cerebus]

How would this "American propaganda" about the chinese Supermicro server backdoor benefit the USA?

To gain support for any war (cold or hot) you need the populace to believe that the "enemy" is threatening you.   Why not baffle them with mysterious electronics jargon and a magical chip "the size of a grain of rice" that allows them to spy on us. (Almost as good as the omnipotent Russian hackers and social media trolls)

Geopolitics 101 - same as it ever was.  There needs to be more than just a trade war to justify military spending.

A new cold war with China is in the making.

I'll let a better man than me speak:

    A vital element in keeping the peace is our military establishment. Our arms must be mighty, ready for instant action, so that no potential aggressor may be tempted to risk his own destruction...

    This conjunction of an immense military establishment and a large arms industry is new in the American experience. The total influence—economic, political, even spiritual—is felt in every city, every statehouse, every office of the federal government. We recognize the imperative need for this development. Yet we must not fail to comprehend its grave implications. Our toil, resources and livelihood are all involved; so is the very structure of our society. In the councils of government, we must guard against the acquisition of unwarranted influence, whether sought or unsought, by the military–industrial complex. The potential for the disastrous rise of misplaced power exists, and will persist. We must never let the weight of this combination endanger our liberties or democratic processes. We should take nothing for granted. Only an alert and knowledgeable citizenry can compel the proper meshing of the huge industrial and military machinery of defense with our peaceful methods and goals so that security and liberty may prosper together. [emphasis added]

[Cerebus]

Software.

You can relatively easily find that software/firmware was altered. If it's a tiny innocent looking EMI filter, sleeping until comes time to do it's dirty job, it's way more difficult to figure it out.

You've not read Thompson's paper have you? If you had you would know you were talking out of your hat.

The moral is obvious. You can't trust code that you did not totally create yourself. (Especially code from companies that employ people like me.) No amount of source-level verification or scrutiny will protect you from using untrusted code.

[xaxaxa]

time to buy supermicro stock? 

[JoeO]

I would be surprised if the Chinese were NOT doing this type of activities.

[EEVblog]

Doing this SuperMicro motherboard implant as a component retrofit (chip disguised as a decoupler or ESD protection or whatever) without it being obvious (cuts/bodge wires)
seems quite implausible.

Factory design engineered?

[EEVblog]

One other way to do it would be by the PCB supplier modifying the gerbers to add the part. But then the assembler has to be in on it too.

Supermicro are probably being gagged with a fear of losing contracts or whatever, as it would be easy for them to come out and prove it wasn't in their design and layout, and that's it was some manufacturing  chain interdiction.

[TimNJ]

Doing this SuperMicro motherboard implant as a component retrofit (chip disguised as a decoupler or ESD protection or whatever) without it being obvious (cuts/bodge wires)
seems quite implausible.

Factory design engineered?

Kind of my main point of confusion. Sure, you can make a custom chip to harvest data off some I2C-esque bus. But, the board was (presumably) not designed to accommodate this extra chip to start with...so how'd they get it on there? You can't just stick a chip on a board willy-nilly...obviously..

The photo they showed (assuming that's really it) looked like a directional coupler. Something like this one: https://product.tdk.com/info/en/products/rf/rf/coupler/index.html

Is it possible that Supermicro had made accommodations for a directional coupler (or whatever it is), and then decided not to populate it? Even if that's true, couplers are passive parts; There shouldn't be a power supply connection.

Or, perhaps, the board was modified at the PCB manufacturer/assembly house. I suppose they could have went in, modified the Gerbers, and added one more part to the PnP list...But if that's the case, then Supermicro should be able to come out and say "Hey that PCB layout was never approved by us!". But they haven't said that.

Or Supermicro is directly colluding with China.

[TimNJ]

One other way to do it would be by the PCB supplier modifying the gerbers to add the part. But then the assembler has to be in on it too.

Supermicro are probably being gagged with a fear of losing contracts or whatever, as it would be easy for them to come out and prove it wasn't in their design and layout, and that's it was some manufacturing  chain interdiction.

Exactly my thoughts.

[JustMeHere]

I would have my SSL handshake hide data in the session setup.  You would not be easily able to detect it.  When the system does an online BIOS update, send the data. 

For example the N position in the prime number is actually byte data.  Have N change every time.