Fedora May Seek MS Code Signing for Secure Boot

CommanderFrank

Cat Can't Scratch It
Joined
May 9, 2000
Messages
75,399
A compromise solution proposed by a Fedora developer could have versions of Fedora come with a bootloader signed by Microsoft. The move would guarantee Linux distros ease of installation on computers with the secure boot mechanism.

Although that trend is distressing to sophisticated Linux enthusiasts who value technical freedom and autonomy, the clear security advantages are a boon to regular end users who need all the help that they can get to minimize the threat landscape.
 
Hopefully a better solution will be found very soon: I don't like the idea that only Microsoft certificates appear to be valid by default...not only are there other parties whose signatures can/should be trusted, but I specifically do NOT trust Microsoft. I'm not concerned by petty security threats like teenage punks and credit card scammers. I'm concerned by serious security threats, like government and corporate back-doors....and anything written by Microsoft and "guaranteed secure" might as well be "guaranteed to have a built-in back door for spooks."

I'm glad secure boot can be disabled, and it's all well and good that Microsoft is signing a bootloader shim for GRUB - for now - but this creates a dangerous dependence on Microsoft for secure boot. If the ability to disable secure boot disappears over time and Microsoft remains the sole trusted authority, we could run into some real problems with corporate lockdown of PC hardware. After all, this lockdown is already going to be in place with ARM devices, where secure boot cannot be disabled, and the walled garden is inescapable...I worry that this may happen to x86 in the future as well. We seem to be taking gradual steps toward that kind of future right now.

Moreover, the restrictions on signing "privileges" seem ominous as hell: If your Linux distro has enough low-level hardware access that it can tamper with the firmware or Windows boot, then Fedora's own signing privileges get revoked. This essentially means that only a deliberately gimped operating system is allowed to run in secure boot mode. That's aside from the problems with third party drivers...

chris__ also makes a good point in the comments: What if you run a Windows VM inside of Fedora (or any other OS), where the boot process seamlessly launches the VM? In this case, your "Windows" installation is not a Windows installation at all, and so it is therefore compromised. Following from this, the implication is that any OS capable of seamlessly hosting another is unfit for being considered "secure," and must not run in secure boot mode.
 
Gee, and here I was thinking penguins spent all their time marching. THE MOVIE LIED TO MEEE!!!
 
(Note: When I refer to an "inescapable" walled garden above, I'm excluding jailbreaking methods that may become more difficult to do in the future, either technically or legally.)
 
....and anything written by Microsoft and "guaranteed secure" might as well be "guaranteed to have a built-in back door for spooks."

Not to spoil the seriousness of your post, but it would have been funnier if you would have said either, "a built-in back door for sporks," or, "a built-in back door for Spocks."
 
chris__ also makes a good point in the comments: What if you run a Windows VM inside of Fedora (or any other OS), where the boot process seamlessly launches the VM? In this case, your "Windows" installation is not a Windows installation at all, and so it is therefore compromised. Following from this, the implication is that any OS capable of seamlessly hosting another is unfit for being considered "secure," and must not run in secure boot mode.

This is a pretty good point and it also makes me question the ramifications of VM's of Windows in general .
 
For me secure boot won't be an issue, like many HardOCP people here. The problem is other people, the joe six packs. If you want to persuade people to use an alternative OS, secure boot is going to screw that all up. Look how much effort Ubuntu put into their OS for any Windows user to try it out. The CD is totally automated and will do everything for you, including partitioning the hard drive if you choose to go that way. Now try to convince people to hit a cumbersome array of keys to enter a menu and navigate it to disable secure boot. It's just not going to happen.
 
Wait, back the truck up.

Somehow Microsoft has wrestled away control of how, and what, a PC can boot?

Am I the only one that finds this a horrific idea?
 
Wait, back the truck up.

Somehow Microsoft has wrestled away control of how, and what, a PC can boot?

Am I the only one that finds this a horrific idea?

Not at all, and it's a great idea.

By running secure boot, you ensure that only digitally signed bootloaders are able to start when you power up your system. This will eliminate viruses that modify your MBR and are able to compromise your system before you can even start it and drop into safe mode.

When I was at HP, we were in the process of getting all the firmwares in ProLiants moved to enforced digitally signed firmware. It's a security measure required for some levels of FIPS certification and is soon going to be a requirement for DoD contracts.
 
Wait, back the truck up.

Somehow Microsoft has wrestled away control of how, and what, a PC can boot?

Am I the only one that finds this a horrific idea?

That's not completely true. When secure boot is enabled, the system requires that all firmware and software in the boot process has a certificate signed a trusted CA. It's sort of an oversimplification, but a Linux distro really just needs to obtain the necessary certificate and present it during OS start. Atop that, you should be able to turn secure boot off in the BIOS so another OS can be used.

I haven't read up on it so much that I'm anything close to an expert, but it'd be a little preemptive to assume that the only trusted CAs are Microsoft owned. Furthermore, MS has become a little more friendly to the open source community in recent years. Serious attempts to shoulder out other operating systems from the market might land them in another antitrust lawsuit. They already have a monopoly and while that's not illegal, it is only a few short steps off the line to give the public the impression they're using the monopoly in a manner that can land them in court.
 
You'd think that this would have anti-trust and anti-competitive business practices written all over it. A conspiracy between MS and hardware manufacturers to shutout alternative OS from running on future hardware.
 
reading the article it looks like linux is already working on their own implementation. Unless I read it wrong this was from windows8 pc dual booting to a linux distro. Seriously would an real linux geek do that? That would probably just VM the windows 8 os ;)
 
reading the article it looks like linux is already working on their own implementation. Unless I read it wrong this was from windows8 pc dual booting to a linux distro. Seriously would an real linux geek do that? That would probably just VM the windows 8 os ;)

In practical terms, I actually think more Linux users would want to dual-boot Windows rather than host it: If Linux is your primary desktop, then you usually only need Windows for a select few pieces of software that won't run (or run well) in Wine. Typically, that means games - which VM's would need more native graphics hardware access to run well - or programs that interface specialized hardware without Linux drivers (e.g. capture cards), which would also be unavailable through a VM.

Of course, this post is off topic, but...
 
All I see is a lot of FOSS people throwing FUD to make it look like Microsoft is monopolizing the boot process when the main purpose is to prevent unauthorized malicious code to launch before the OS can do anything.

User Workarounds:
1) If you're using a popular linux distro (like Fedora, Ubuntu, or SLED), just have them get a certificate. It looks like the direction they are going.
2) If you want to run your own custom build, then disable it in the BIOS. I find it extremely hard to believe people who compile their own kernels can't navigate a BIOS.
 
reading the article it looks like linux is already working on their own implementation. Unless I read it wrong this was from windows8 pc dual booting to a linux distro. Seriously would an real linux geek do that? That would probably just VM the windows 8 os ;)

It depends on your workload. If you are doing computational/graphics/IO intensive stuff in Linux, then it would be best to have Windows in a VM. If it's vice versa, then you should have Linux in a VM.

However, given a very typical use case scenario of developing on Linux and playing 3d games on Windows, the Linux VM inside Windows is probably the optimal path for people (Windows 8 has built-in Hyper-V).
 
(Windows 8 has built-in Hyper-V).

WTFBBQSAUCE?! That's fantastic and I really should have read the blog about it from last September. Oh evil rock, why must it be so comfortable to live under you?
 
All I see is a lot of FOSS people throwing FUD to make it look like Microsoft is monopolizing the boot process when the main purpose is to prevent unauthorized malicious code to launch before the OS can do anything.

They do that with any security measure that could be constructed to sound like DRM.

Requiring signed code in order to prevent malicious modification of your bootloader turns into an article on Slashdot titled "Microsoft working with motherboard makers to shut out Linux!"
 
The problem is, if you read any of the better Linux Dev Blogs, UEFI really IS a huge problem. You say "just go buy a cert" ... I do not care if it is Win8 or Linux for a moderately heavy desktop pc user like me that is going to be a big pain in the ass. Now for purely Corporate purposes it would likely be heaven on earth.
 
The problem is, if you read any of the better Linux Dev Blogs, UEFI really IS a huge problem. You say "just go buy a cert" ... I do not care if it is Win8 or Linux for a moderately heavy desktop pc user like me that is going to be a big pain in the ass. Now for purely Corporate purposes it would likely be heaven on earth.

It's the secure boot piece of UEFI rather than UEFI itself that's a concern. As others mentioned, secure boot can be easily disabled simply by turning the feature off so a set of trusted certificates isn't necessary. It seems like it's more or less just a minor PITA rather than something that will prevent Linux distros from running on a machine.
 
The problem is, if you read any of the better Linux Dev Blogs, UEFI really IS a huge problem. You say "just go buy a cert" ... I do not care if it is Win8 or Linux for a moderately heavy desktop pc user like me that is going to be a big pain in the ass. Now for purely Corporate purposes it would likely be heaven on earth.

How is it a pain for you whether or not a certain Linux distro signs their boot loader? The big distros will undoubtly get it signed. If you're using a small or custom build, you go to BIOS and disable secure boot. This takes less than 1 minute and you never have to do it again.

You have to understand the background where FOSS and Microsoft comes from. FOSS mainly cares about compatiblilty and features. Microsoft cares about security, due to their less than glorious past and public perception. I was rather surprised initially, when I began working with both FOSS and Microsoft on a certain hardware feature. Microsoft was literally 2 years late to market for OS support, but they found gaping security holes that the open source community neglected.

WTFBBQSAUCE?! That's fantastic and I really should have read the blog about it from last September. Oh evil rock, why must it be so comfortable to live under you?

Yes it is full blown Hyper-V. The only relevant virtualization feature missing compared to Server 2012 is SR-IOV.
 
How is it a pain for you whether or not a certain Linux distro signs their boot loader? The big distros will undoubtly get it signed. If you're using a small or custom build, you go to BIOS and disable secure boot. This takes less than 1 minute and you never have to do it again.

Yes, until the ability to turn it off is removed in future x86 hardware, like it will be in ARM hardware from the start.
 
Yes, until the ability to turn it off is removed in future x86 hardware, like it will be in ARM hardware from the start.

Precisely, there is no guarantee that there will be a way of turning it off.
 
Yes, until the ability to turn it off is removed in future x86 hardware, like it will be in ARM hardware from the start.

First of all, that's pure conspiracy theory.

Second, even if it did ever happen, no one except the ideologists would be crying. It's a great security feature, and every operation that's even half-legit would spend $20 on a certificate. How often do you have a critical need to patch GRUB that would result in catastrophic failure if you had to submit the patch upstream and wait for the next signed nightly build?

The only people who will ever need to run unsigned bootloader are firmware programmer, and they have unlocked development boards.

This whole ordeal is nothing more than more rampant speculation from the FOSS crowd, fueled by the usual conspiracy theories that everyone in the for-profit tech industry is working together to squash the free software movement.
 
This whole ordeal is nothing more than more rampant speculation from the FOSS crowd, fueled by the usual conspiracy theories that everyone in the for-profit tech industry is working together to squash the free software movement.

That's true. The only tech industry company that wants to turn Open Source into the next dead quadrotor cat is Oracle.
 
First of all, that's pure conspiracy theory.
Yes because MS has never ever done anything at all to be anitcompetitive. That and Apple's Itunes is merely a method of saying "I love you". :rolleyes:

Second, even if it did ever happen, no one except the ideologists would be crying. It's a great security feature, and every operation that's even half-legit would spend $20 on a certificate. How often do you have a critical need to patch GRUB that would result in catastrophic failure if you had to submit the patch upstream and wait for the next signed nightly build?
Secure boot not only affects bootloaders, but kernels, drivers etc. Any device that is brought online by UEFI and later accessed by the kernel must have drivers that are not directly accessible by the end user. For example to my knowledge Nvidia doesn't offer drivers for Linux that support KMS so until they do... well no booting for you.

The second problem is that SecureBoot doesn't solve much of anything in the grand scheme of things if you think about it. When it comes to security, if you have an Internet connection then you are at risk... period...full stop. Are you going to replace any other machines on your network that don't have UEFI? Your file server? Your HTPC? Your tablet? Your phone? If not then they are nodes that could be used to attack you. What's the effectiveness of your firewall? Don't know? Surely you would know whether or not the kernel which has all of those drivers in kernel space is fully secure or not...no? Ok maybe not that....well surely your browser clears all history, cookies any and all files associated with pages you've visted even the cookies that exist in tabs that have opened along side of this page you are viewing, no? Hmm well I guess be glad you've got SecureBoot keeping your BIOS safe while all of the viruses execute in user space.... you know where your banking information is.

The only people who will ever need to run unsigned bootloader are firmware programmer, and they have unlocked development boards.
Um without the driver, just dealing with firmware isn't going to get you far.

This whole ordeal is nothing more than more rampant speculation from the FOSS crowd, fueled by the usual conspiracy theories that everyone in the for-profit tech industry is working together to squash the free software movement.
There's nothing speculative about it. MS has a long long history of being anti-competitive. Normal people would be weary of dealing with someone with MS's track record. Hey how's Nokia doing BTW?
 

I don't recall Microsoft being in control of the manufacture of motherboards with UEFI. You're assuming that a company producing hardware is going to artifically limit their potential market by forcing Secure Boot, but that doesn't make economic sense. There'd have to be some heavy incentives on the table and, even if there were, it's evident in this thread that Linux distros can obtain a signed certificate from a trusted CA.
 
I don't recall Microsoft being in control of the manufacture of motherboards with UEFI. You're assuming that a company producing hardware is going to artifically limit their potential market by forcing Secure Boot, but that doesn't make economic sense. There'd have to be some heavy incentives on the table and, even if there were, it's evident in this thread that Linux distros can obtain a signed certificate.

UEFI boards have been around for years at this point, but Secure Boot as a feature has basically gone unused on the desktop.

Now who is requiring its use? Is Hp? No. Is The linux Foundation? Nope. How about Dell? Naw dont think so. So who Is? I believe that is Microsoft. Microsoft can control what the Mb manufacturers do because Microsoft controls at least 90% of the global market of all operating systems installed within the US. Anyone who thinks they cant is smoking some serious shit. If that person is you, I think you should share. Dont be stingy.

With regards to FOSS buying thier own certificate, why should they have to do that?
 
UEFI boards have been around for years at this point, but Secure Boot as a feature has basically gone unused on the desktop.

Now who is requiring its use? Is Hp? No. Is The linux Foundation? Nope. How about Dell? Naw dont think so. So who Is? I believe that is Microsoft. Microsoft can control what the Mb manufacturers do because Microsoft controls at least 90% of the global market of all operating systems installed within the US. Anyone who thinks they cant is smoking some serious shit. If that person is you, I think you should share. Dont be stingy.

You seem upset. I don't think much of this doom and gloom is motivated by rational thought. On the off chance that Linux ceases to exist (which it won't since several vendors are already making moves to obtain certificates) the world really hasn't lost much of anything. Don't get me wrong, I love my Hannah Montana Linux box and find it useful to keep a copy of Mint and Puppy available, but if Secure Boot suddenly vaporizes the entire FOSS community, hardly anyone will even notice it vanished.

With regards to FOSS buying thier own certificate, why should they have to do that?

If they want to maintain relevance in a world incresingly concerned about security (not the illusion of security through obscurity that Linux presents) then they've got to keep up with the companies that are enjoying the comfortable luxury of being able to set those trends. The reality, like it or not, is that FOSS represents a small segment of enterprise and personal software. While lots of us like cheering for the underdog, the underdog has no choice but to follow the market leaders. Ultimately, I think most of the outrage isn't about Secure Boot, but about how the implementation, something Microsoft did in a casual, dismissive manner, highlights how little relevance FOSS has in the world. Pointing out something as having no influence is ego crushing for champions of Open Source and their only outlet is to cry foul to a world that isn't really listening.
 
Yes because MS has never ever done anything at all to be anitcompetitive. That and Apple's Itunes is merely a method of saying "I love you". :rolleyes:

....

There's nothing speculative about it. MS has a long long history of being anti-competitive. Normal people would be weary of dealing with someone with MS's track record. Hey how's Nokia doing BTW?

I'm not responding to these, there's no point arguing with crazy and baseless conspiracy theories..

Secure boot not only affects bootloaders, but kernels, drivers etc. Any device that is brought online by UEFI and later accessed by the kernel must have drivers that are not directly accessible by the end user. For example to my knowledge Nvidia doesn't offer drivers for Linux that support KMS so until they do... well no booting for you.

THAT'S A GOOD THING. The easiest place for a virus to hit is hooking in at the low level. SecuROM and all the DRM schemes do this. Remember Sony's rootkit fiasco? That too.

Requiring the entire chain of command to be digitally verified by each device's manufacturer ensures that a malicious attacker will never be able modify say your network and keyboard drivers and start keylogging everything.

This is built-in protection against not only malicious viruses, but also illegal government wiretapping (yes, the FBI has been caught before spreading viruses to subjects of investigation for monitoring purposes).

The second problem is that SecureBoot doesn't solve much of anything in the grand scheme of things if you think about it. When it comes to security, if you have an Internet connection then you are at risk... period...full stop. Are you going to replace any other machines on your network that don't have UEFI? Your file server? Your HTPC? Your tablet? Your phone? If not then they are nodes that could be used to attack you. What's the effectiveness of your firewall? Don't know? Surely you would know whether or not the kernel which has all of those drivers in kernel space is fully secure or not...no? Ok maybe not that....well surely your browser clears all history, cookies any and all files associated with pages you've visted even the cookies that exist in tabs that have opened along side of this page you are viewing, no? Hmm well I guess be glad you've got SecureBoot keeping your BIOS safe while all of the viruses execute in user space.... you know where your banking information is.

Basically you're trying to dismiss low-level attacks as some sort of irrelevant thing because there's lots of high-level attacks, which is just stupid.

Low-level attacks are terrifying because they are undetectable and there's no way to know they are cleaned off. The only way to be sure you've cleared a bootloader virus is to boot from another device and overwrite your entire bootloader. If it was a BIOS-level attack, you can't be sure that the flashing mechanism wasn't compromised too; literally the only way to verify it's gone would be to pull your BIOS chip and put it on a EEPROM programmer. These kind of attacks are in the works and quickly coming down the pipeline, primarily from the Asian countries.

When I was at HP we actually came up with a way to flash the BIOS on a server through the iLO, which is a network-connected device. All of the security measures you listed are useless then - once you've got the BIOS, you own the server.
 
First of all, that's pure conspiracy theory.
Considering Microsoft past actions, this would not be surprising at all.

Second, even if it did ever happen, no one except the ideologists would be crying..
What the fucking fuck? Not being able to run what you want on a computer (which is why it is a computer) is "ideologist"?

As always, it looks like the only ones accusing others of extremism are the crazy extremists.
 
First of all, that's pure conspiracy theory.

Second, even if it did ever happen, no one except the ideologists would be crying.
Hahahahaha. I love when people do this, and it happens all the time. Someone tries to dismiss someone else's concerns by saying, "Oh, this is a silly conspiracy theory, and it's never going to happen, blah blah," but they're fully aware that it's probably just a matter of time, so they include an escape clause, "...and if it did, it would be a great thing, and I would be fully in support of it, and I hope it happens...but you know, this is still totally a conspiracy theory, and there's no way on earth anyone else in a position of power who stands to benefit from control would share my point of view, so people who disagree with me have nothing to worry about."

LO-f'ing-L.

Oh, and, "Only ideologists would be crying?" If anyone concerned about unchecked third party control over what THEIR computer hardware can boot is an ideologist, then yes. I'll go with the flow when it makes sense to me, but I make no apologies for reserving the right to demand full control over my hardware.

It's a great security feature, and every operation that's even half-legit would spend $20 on a certificate. How often do you have a critical need to patch GRUB that would result in catastrophic failure if you had to submit the patch upstream and wait for the next signed nightly build?

The only people who will ever need to run unsigned bootloader are firmware programmer, and they have unlocked development boards.

I don't think you understand what the article is saying about the requirements imposed on operating systems for their authors (or maintainers, or whoever) to be eligible for a signing certificate for secure boot. You can't just sign any old bootloader or bootloader shim to verify it hasn't been tampered with. The OS has to be gimped on purpose, as well.

Article said:
In order for the technical advantages of the secure boot mechanism to be fully realized, however, all of the code in the platform that has direct interaction with hardware has to be trustworthy, too. A malicious party could theoretically use the Linux kernel’s low-level hardware access to compromise a Windows installation on the same computer or tamper with the firmware.

If that is possible on a Fedora system with a signed bootloader, then Fedora’s signing privileges would be revoked and the operating system would no longer be able to run in a secure boot environment.
Much of that is possible on Linux. Even creating a Windows VM - especially one that seamlessly starts upon Fedora booting, without the user knowing Fedora is being run - could fall afoul of this requirement.

Article said:
To prevent such a scenario from occurring, Fedora will set up its own signing system that will be applied to the kernel and other security-sensitive layers of the stack below the Microsoft-signed bootloader initialization layer.
...except simply signing won't be enough, because first you have to gimp the code, then sign it. Then, add in difficulties with third party device drivers, and secure boot poses a whole ton of problems for Linux.

Of course, you can always turn secure boot off to avoid these problems...until you can't, like you seem to imply would be a good thing.

This whole ordeal is nothing more than more rampant speculation from the FOSS crowd, fueled by the usual conspiracy theories that everyone in the for-profit tech industry is working together to squash the free software movement.

Would Microsoft like to undermine Linux with tactics like this? Well, yeah, of course they would. Embrace, extend, extinguish has been their motto in the past, and they have a long track record of belligerent anti-consumer behavior to back it up.

There is more going on here though. It is no secret that the US government is hellbent on using every possible consumer appliance to spy on citizens. They say so explicitly, but to talk about it is "conspiracy theory." The closer we get to a walled garden on x86 hardware, the closer we get to only being able to run a deliberately compromised system, in the interests of "national security." (Of course, I wouldn't be surprised if I heard a litany of BS like, "Well, they already spy on your searches, and blah blah fricking blah, so why not just add to the problems?" The last part, about adding to the problems, succinctly describes my thoughts on such a submissive point of view.)
 
...except simply signing won't be enough, because first you have to gimp the code, then sign it.

How is incorporating the presentation of a trusted certificate to the firmware going to limit the functionality of the software presenting the certificate? Present the cert, get a nod from the system, then continue processing.
 
How is incorporating the presentation of a trusted certificate to the firmware going to limit the functionality of the software presenting the certificate? Present the cert, get a nod from the system, then continue processing.

I'll requote part of the article:

In order for the technical advantages of the secure boot mechanism to be fully realized, however, all of the code in the platform that has direct interaction with hardware has to be trustworthy, too. A malicious party could theoretically use the Linux kernel’s low-level hardware access to compromise a Windows installation on the same computer or tamper with the firmware.

If that is possible on a Fedora system with a signed bootloader, then Fedora’s signing privileges would be revoked and the operating system would no longer be able to run in a secure boot environment.

If the entire OS isn't gimped, Fedora's signing "privileges" will be revoked.
 
I'll requote part of the article:

If the entire OS isn't gimped, Fedora's signing "privileges" will be revoked.

Ah, I think I understand where you're going with this. If the OS kernel, using its ability to access hardware directly, allows Linux to be used to go tamper with a Windows install, then the distro gets its cert revoked because it poses a security risk. Preventing that level of access is the gimping you're talking about, right?

Keeping in mind that I'm in this discussion deeper than I usually go so I admit that I'm out of my league...if that's correct, then wouldn't any OS be subject to the same possible conditions? At some point in the software chain, stuff has to be able to interact with the metal that makes it go. What is it that makes Windows any more special and immune to low-level hardware access?
 
Ah, I think I understand where you're going with this. If the OS kernel, using its ability to access hardware directly, allows Linux to be used to go tamper with a Windows install, then the distro gets its cert revoked because it poses a security risk. Preventing that level of access is the gimping you're talking about, right?

Keeping in mind that I'm in this discussion deeper than I usually go so I admit that I'm out of my league...if that's correct, then wouldn't any OS be subject to the same possible conditions? At some point in the software chain, stuff has to be able to interact with the metal that makes it go. What is it that makes Windows any more special and immune to low-level hardware access?

The specifics here are very murky, because it's unclear exactly what is meant by modifying a Windows install: Technically, anyone with unhindered write access to a hard drive could do this...but if the entire Windows installation (not just the boot loader) is digitally signed and verified, any tampering could be detected at boot. Strictly speaking, no possible OS - no matter the root user's level of hardware access - could directly tamper with such an install unless the attacker had access to the signing key.

At the same time, the article refers to the potential of Fedora being able to do so, in which case Fedora's signing privileges would be revoked. There's an apparent contradiction here, so either the author of the article is confused, or the restrictions go deeper. For instance, if an OS allows the root user to access the UEFI and turn secure boot off, a Windows partition could be completely overwritten with a compromised installation...therefore, the authoritarian solution would be to disallow any OS from allowing any user (even root) to turn off secure boot once an OS has been loaded. This is just a hop, skip, and a jump away from making secure boot mandatory altogether.

By design, Linux gives the root user (and user code running as root) a lot more control over the operating system and the hardware than Windows does. The Windows kernel code has full hardware access of course, but even administrative users are unable to wield this power as their own. Depending on how restrictions on signing "privileges" are interpreted - and as the infrastructure becomes more ubiquitous, the inertia behind the technology will allow certain players to tighten their grip - there's a potential that any operating system like Linux that acknowledges the user's ownership of their hardware (and gives them the tools to access it) would no longer be permitted.

To expand on another security threat: If you take the ability to host virtual machines to its logical conclusion, a host OS can act like a transparent hypervisor. If such an OS is capable of hosting a modified Windows OS, then the implication is that hypothetically, a hacker who gained root access on any system (Windows, Linux, or otherwise) might install a hosting-capable OS like Linux on a user's machine, install a compromised Windows VM on that host OS, and configure the host to immediately launch the compromised VM in fullscreen at boot, then reboot to the new installation, without the user ever knowing they aren't running their original Windows installation. This elaborate scheme is obviously a security threat (which depends upon another vulnerability giving a hacker root access in the first place), but if the answer is, "Make it impossible for 'allowed' operating systems to host virtual machines," or, "Make it impossible for 'allowed' operating systems to host virtual machines unless the VM is also signed and approved by Big Brother," then we have a problem.

Of course, any hypothetical restrictions to VM host capabilities would apply to Windows as well as Linux...unless, you know, someone with enough clout pulled enough strings to make an exception. :rolleyes: Therein lies the central problem with this signing scheme: It creates a chain of trust and control that includes corporations and perhaps government at some point, but it does not include the user who actually owns the hardware. If secure boot becomes mandatory for x86 - and there will be increasing corporate/government pressure to make it so (under thinly veiled pretexts easily swallowed by the gullible) - it will strip the user of their ability to make their own decisions about what code to run on their hardware. It relies on the assumption that certificates stored in firmware will always and forever be issued/allowed either freely or on the basis of merit...yet the very nature of centralized (or nearly centralized) control over what certificates are permitted and not, gives that authority WAY too much arbitrary power over whose code is allowed to be run and whose is not...on your hardware.

Moreover, the Linux kernel itself is modifiable by advanced users who compile the kernel directly from source and install optional patches to change functionality (e.g. Con Kolivas's performance patches of yore). Obviously, self-compiled code will not be signed with Fedora's signing certificate. Most Linux users do NOT do this, because it's a huge pain, but the ability to do so is integral to open source software: Linux is not just a binary blob packaged by a vendor; users always have access to the source code, so they can compile it themselves if any particular vendor's trustworthiness (or the trustworthiness of all of them) ever comes into question. The freedom to do so is not only helpful to programmers and tinkerers in an everyday sense, but it's crucial for everyone in a political sense: The convenience of a packaged solution usually warrants placing your trust in the vendor, but the fallback ability to "roll your own" or install "underground," non-corporate software is an important safeguard against a world where all of the "legitimate," Big Brother-approved software might be coming installed with back doors.

At their full potential, Linux, the BSD's, and open source software itself are more than just geek toys and server software, and software freedom is about more than vanity and ideology. In a worst case political scenario of corporate vendors colluding with a totalitarian government, FOSS can and will assume the role of "underground" technological life preservers. The ability to compile software from source yourself and run it is a crucial check against corporate and governmental control over all computing and communications technology, but it disappears in a world where the firmware only allows corporate-signed bootloaders to boot, and a corporation's signing authority relies on the entire OS being up to some recognized authority's "standards."

We are entering a world where dystopian police states like Orwell's 1984 have become technologically feasible, and we are not only well on our way politically, but we're cheering on the construction of its ironclad technological infrastructure in the name of temporary security from petty threats like Anonymous and identity thieves. It is foolish to place all of your eggs in the basket that reads, "It'll never happen here, and there will never again for the rest of mankind be need for revolution." The entire history of human civilization is a history of governments and their beneficiaries getting completely out of control, and governments in the 20th century alone killed over 100 million (or is it 200 million?) of their own people.

Secure boot may be extremely useful for many scenarios (including business computers, etc.), but we live in a world with very few motherboard and firmware manufacturers, and we have not yet reached the point where the system BIOS (or UEFI) can be easily replaced by an open source equivalent like Coreboot. In fact, secure boot aims to make it HARDER to tamper with the firmware, since the ability to flash the firmware may have to be removed from Linux just so Fedora et al can retain signing privileges. This kind of technology has already been used by vendors to lock phone users into a walled garden, and secure boot will be used in exactly that manner from the start on ARM devices, so I have VERY little confidence that an ability to disable it on x86 will last forever. Once this technology really takes hold and gains momentum, it will be all too easy for the ability to disable secure boot to be removed entirely, and I'm not exactly keen on taking this risk until an easy fallback solution exists for a user to retain control over their computer. That's why I'm trying to make people aware of the dangers of this technology...if enough people are concerned enough to generate friction, it will help keep alternatives alive and slow down a potentially reckless adoption pace.
 
The specifics here are very murky...

Holy marathon post! :eek:

Just a couple of thoughts, it'd be rather difficult to insert a hypervisor and fully configured OS into a machine without having physical control of at least the primary hard disk. If that's the case, then there's probably an insider threat and that has bigger implications that Secure Boot really won't do anything to solve.

I don't see why integration of a certificate is outside the realm of possiblity for custom compiled kernels or to account for the exceptional circumstance. As you've said, this isn't something people really do that often. I'm sure there are scientific applications or clustered systems being asked to do off-the-wall work that get that sort of treatment, but for turnkey desktop OS builds, that's just not necessary. (Please don't murder me in my sleep Gentoo users!) The last time I even bothered with such a thing was in a Slackware build like 10 years ago and that was just to say I'd done it.

I guess the ultimate question still lurking is whether or not Secure Boot will eventually become a mandatory feature. This is the point where I think Microsoft is going to have to tread carefully (if it's pushing for that very thing). If it becomes clear that mandatory Secure Boot is an anti-competitive practice that results in other OS platforms being unable to function, that might attract the DoJ's attention UNLESS (get your tinfoil hat) it's acting with the approval of a government agency.
 
Okay, one last thought:

Why don't these interested Linux parties actually join the UEFI standards body and take an active role in developing UEFI (including Secure Boot) so that it doesn't screw them over?

http://www.uefi.org/join/

If the FOSS community took a participating interest in their own survival, it might be to their benefit in the long run.
 
Back
Top