5 ms·

New graphics engine imperils users of Firefox and Chrome

20 points by fvbock 15y ago

wccrawford 15y ago
Imperils? No. Leaves another possible angle of attack for hackers? Sure. But then, every feature does. If you want to be safe, use Links on OpenBSD. If you want to browse the web in style, use Chrome and Firefox.
- IDisposableHero 15y ago
  every feature ... leaves another possible angle of attack for hackers Hm, I think the point is that some features are worse than others (larger attack surface, more bugs exposed), and WebGL is particularly bad.
- palish 15y ago
  Erm... "Lynx" whistles innocently
  - wccrawford 15y ago
    Okay, I'll bite. Why is Lynx better than Links? I was under the impression that Links was the newer one, though I hadn't done much research.
    - palish 15y ago
      I had no idea that there was a "Links". I thought he meant "Lynx", but misspelled it. Now I want to write an operating system and name it "Debyan".
stianan 15y ago
Could this have anything to do with VUPEN's recent Chrome exploit?
- blocke 15y ago
  "Could this have anything to do with VUPEN's recent Chrome exploit?" Is this going to be the standard post for every Chrome fix from this point forward? The only people who we know for sure would know ain't talking. ;)
  - stianan 15y ago
    I get your point. However, this isn't a fix, it is, according to the article, a volunerability that allows "remote execution of malicious code". People who know WebGL better than I do would probably know more about what specific exploits it facilitates.
    - JoeAltmaier 15y ago
      That's pretty much every attack - the holy grail of malicious code.
- evangineer 15y ago
  It could be. Who knows. What I do know is that Chrome & ChromeOS have got some of the most diligent security-minded hackers in the world working on them. I get occasional updates on the stuff they are doing to make Chrome secure. Things like the sandbox, SPDY, HSTS and more. This sort of thing will just motivate them to work smarter & harder.
prewett 15y ago
GL_ARB_robustness? I thought this might be satire or something, but it actually exists: http://www.opengl.org/registry/specs/ARB/robustness.txt Apparently it does the same thing for OpenGL that strncpy does for strcpy.
- palish 15y ago
  You mean strcpy_s?
  - owenmarshall 15y ago
    OpenBSD po-ta-to, Windows po-tah-to.
    - palish 15y ago
      It was tongue-in-cheek. I think it's silly that Microsoft adds "_s" to a lot of the standard functions; it makes porting more difficult. Of course, it was sillier not to write "strcpy" as strcpy(dest,maxSize,src) from the beginning. And "memcpy" as memcpy(dest,maxSize,count,src). But I guess whoever wrote those weren't really expecting them to be one of the most frequently used functions in computer programming history...
extension 15y ago
I'm not surprised. Video drivers are big, complicated and buggy. They are developed within very tight performance and time constraints and have had little reason to think about security until now. Exposing a large part of them to the public internet is begging for trouble. On the other hand, if it's ever going to happen, it will probably take a few embarassing public incidents to make the vendors change their priorities.
- imurray 15y ago
  Agreed (at least to the first part). The same argument has been made about the new web font technologies. http://hackademix.net/2010/03/24/why-noscript-blocks-web-fonts/ http://hackademix.net/2010/03/24/why-noscript-blocks-web-fon... "It really worries me that the FreeType font library is now being made to accept untrusted content from the web. The library probably wasn’t written under the assumption that it would be fed much more than local fonts from trusted vendors who are already installing arbitrary executable on a computer, and it’s already had a handful of vulnerabilities found in it shortly after it first saw use in Firefox."
- 0x0 15y ago
  This should come as no surprise to anyone, really. 10 years ago we had similar issues with 2d graphics. For example, rendering huge fonts could cause the X display server to crash: http://xforce.iss.net/xforce/xfdb/9313 http://xforce.iss.net/xforce/xfdb/9313 With WebGL, the attack surface is many times larger. I'm sure we'll see plenty of exploits leveraging vulnerabilities in the display subsystem.
code_duck 15y ago
"In Chrome, get to the command line of your operating system and add the --disable-webgl flag to the Chrome command. " Wow, this should do a nice job of confusing everyone.
chadaustin 15y ago
This is a _good_ thing. Graphics drivers are so ridiculously terrible that the pressure to fix them needs to come from somewhere. IMVU's top reported crash is access violations inside of Direct3D. Mass-market 3D is a pain in the ass because of this situation, and Intel, Microsoft, nvidia, and amd aren't doing anything about it.

20 points by fvbock 15y ago

Imperils? No. Leaves another possible angle of attack for hackers? Sure. But then, every feature does. If you want to be safe, use Links on OpenBSD. If you want to browse the web in style, use Chrome and Firefox.

IDisposableHero 15y ago
every feature ... leaves another possible angle of attack for hackers Hm, I think the point is that some features are worse than others (larger attack surface, more bugs exposed), and WebGL is particularly bad.
palish 15y ago
Erm... "Lynx" whistles innocently
- wccrawford 15y ago
  Okay, I'll bite. Why is Lynx better than Links? I was under the impression that Links was the newer one, though I hadn't done much research.
  - palish 15y ago
    I had no idea that there was a "Links". I thought he meant "Lynx", but misspelled it. Now I want to write an operating system and name it "Debyan".

stianan 15y ago

Could this have anything to do with VUPEN's recent Chrome exploit?

blocke 15y ago
"Could this have anything to do with VUPEN's recent Chrome exploit?" Is this going to be the standard post for every Chrome fix from this point forward? The only people who we know for sure would know ain't talking. ;)
- stianan 15y ago
  I get your point. However, this isn't a fix, it is, according to the article, a volunerability that allows "remote execution of malicious code". People who know WebGL better than I do would probably know more about what specific exploits it facilitates.
  - JoeAltmaier 15y ago
    That's pretty much every attack - the holy grail of malicious code.
evangineer 15y ago
It could be. Who knows. What I do know is that Chrome & ChromeOS have got some of the most diligent security-minded hackers in the world working on them. I get occasional updates on the stuff they are doing to make Chrome secure. Things like the sandbox, SPDY, HSTS and more. This sort of thing will just motivate them to work smarter & harder.

prewett 15y ago

GL_ARB_robustness? I thought this might be satire or something, but it actually exists: http://www.opengl.org/registry/specs/ARB/robustness.txt Apparently it does the same thing for OpenGL that strncpy does for strcpy.

palish 15y ago
You mean strcpy_s?
- owenmarshall 15y ago
  OpenBSD po-ta-to, Windows po-tah-to.
  - palish 15y ago
    It was tongue-in-cheek. I think it's silly that Microsoft adds "_s" to a lot of the standard functions; it makes porting more difficult. Of course, it was sillier not to write "strcpy" as strcpy(dest,maxSize,src) from the beginning. And "memcpy" as memcpy(dest,maxSize,count,src). But I guess whoever wrote those weren't really expecting them to be one of the most frequently used functions in computer programming history...

extension 15y ago

I'm not surprised. Video drivers are big, complicated and buggy. They are developed within very tight performance and time constraints and have had little reason to think about security until now. Exposing a large part of them to the public internet is begging for trouble. On the other hand, if it's ever going to happen, it will probably take a few embarassing public incidents to make the vendors change their priorities.

imurray 15y ago
Agreed (at least to the first part). The same argument has been made about the new web font technologies. http://hackademix.net/2010/03/24/why-noscript-blocks-web-fonts/ http://hackademix.net/2010/03/24/why-noscript-blocks-web-fon... "It really worries me that the FreeType font library is now being made to accept untrusted content from the web. The library probably wasn’t written under the assumption that it would be fed much more than local fonts from trusted vendors who are already installing arbitrary executable on a computer, and it’s already had a handful of vulnerabilities found in it shortly after it first saw use in Firefox."
0x0 15y ago
This should come as no surprise to anyone, really. 10 years ago we had similar issues with 2d graphics. For example, rendering huge fonts could cause the X display server to crash: http://xforce.iss.net/xforce/xfdb/9313 http://xforce.iss.net/xforce/xfdb/9313 With WebGL, the attack surface is many times larger. I'm sure we'll see plenty of exploits leveraging vulnerabilities in the display subsystem.

code_duck 15y ago

"In Chrome, get to the command line of your operating system and add the --disable-webgl flag to the Chrome command. " Wow, this should do a nice job of confusing everyone.

chadaustin 15y ago

This is a _good_ thing. Graphics drivers are so ridiculously terrible that the pressure to fix them needs to come from somewhere. IMVU's top reported crash is access violations inside of Direct3D. Mass-market 3D is a pain in the ass because of this situation, and Intel, Microsoft, nvidia, and amd aren't doing anything about it.