MLS Hybrids

[FerrusGroup]

Has there been any thought to discussing MLS Hybrids, like NSA NetTop? It could be a good entry point for discussing SELinux, local application boundaries (LABs), and the use of virtualization technologies such as VMware and Virtual PC. Also would be good vetting within the community as to sandboxing on the local PC. I can think of friendly vectors for IPSec and L2TP tunneling for enclaves too. If it's of interest I can throw together a paper before next year.

And yes, I'm new to the forum.

[FerrusGroup]

Google anyone???

A very dry paper on MLS can be found at http://nsi.org/Library/Compsec/sec3.html

The idea of NetTop is not strictly MLS but more of a functional hybrid.
I think the general direction would be to discuss application boundaries and leave all the data tagging and true MLS for another time.

Here are some basics on NetTop.

http://www.nsa.gov/techtrans/techt00011.cfm

Technical Challenge: To create a user workstation computing architecture based primarily on COTS technology that can function seamlessly under varying levels of information assurance requirements while still presenting an easy-to-use interface, and managed COTS environment.

Description: NetTop is the result of a significant and continuing research effort based on the technical challenge referenced above. Broadly, NetTop is a building block-based architecture and associated prototypes, predominately based on COTS, and designed to address a series of different information assurance requirements. As a by-product, the NetTop architecture reduces the physical and environmental footprint issues typically encountered in high-level information assurance solutions.

NetTop incorporates typical COTS user hardware and software found in most offices, schools, and homes. This technology is then combined with an underlying host operating system, virtual machine monitor, virtual network hubs, network encryptors, and a filtering router that allows multiple machine environments to run simultaneously and to access multiple networks all from the same physical platForm.

Additional research has been performed to address issues associated with the use of "thin clients," methods of providing increased assurance levels, and techniques that can provide failure detection.

The benefit of the NetTop architecture is that it removes security functionality from the control of the end-user OS and applications. Important security functions such as communications encryption can be placed in a separate protected environment that cannot be influenced by user software. Similarly, an isolated filtering router function is used to provide protection from rudimentary network attacks. The modularity of the NetTop architecture and the use of standard TCP/IP networking to connect virtual machines facilitates simple replacement or upgrade of individual components.

Demonstration Capability: Several prototype configurations have been created as part of the ongoing research effort.

Potential Commercial Application(s): There are numerous industrial applications for a single workstation that can be used to simultaneously access multiple networks differing in sensitivity levels. Industrial applications also include alternative methods from FIREWALLS for providing information protection, and protected remote access solutions for telecommuting.

http://news.zdnet.com/2100-9595_22-5...ml?legacy=zdnn

NSA attempting to design crack-proof computer
By Robert Lemos ZDNet News January 31, 2001, 4:00 PM PT

Software emulation firm VMware announced it has teamed up with researchers at the National Security Agency to create a nearly crack-proof computer that can place sensitive data in virtual vaults inside the PC.

The concept, assuming it works, would streamline the methods intelligence agencies use to manage data. At present, the NSA--the military surveillance arm of the United States intelligence community--physically separates networks carrying data of a particular classification. For example, top-secret data might be kept on a different computer than data classified merely as sensitive material. Sometimes, in order for a worker to have access to the information they need, up to six different computers can be on a single desk.

That type of security is called--in typical intelligence community jargon--an "air gap." It works, but its days could be numbered, said Ed Bugnion, founder and manager of research and development for Palo Alto, Calif.-based VMware.

"I believe we have a solution out there that provides security comparable to having multiple computers," he said.

Called "NetTop," VMware's answer would turn each computer into a number of virtual PCs running on a Linux computer that would sit on each worker's desk. The security system would erect supposedly impenetrable, but virtual, walls between public data and more sensitive information on the same computer.

If successful, the project could mean huge cost savings and convenience for the NSA and other security-conscious government agencies by eliminating one or more computers--and a variety of network components--cluttering desktops at the agency.

Saving money through software
Paul Pittelli, director of information assurance research at the NSA, said the move is part of the agency's new emphasis on saving money through the use of commercial software.

"Users in the national security community have an increasing need for commercial off-the-shelf software," he said in a statement. "We currently require them to use different computers for different applications." That, Pittelli said, will stop if efforts like NetTop succeed.

VMware's plan is to use an offshoot of the company's current virtual machine technology that allows Linux users to install and run Windows or any other PC-based operating system on top of Linux.

The reason the company believes it can succeed is because it doesn't emulate the software but the hardware underneath.

"Java needs a proprietary environment to run," said VMware's Bugnion. "We can run arbitrary operating systems within the PC." Last year, the company also released a version of its software that runs on Windows NT and 2000, enabling users to run Linux (or any other operating system) in a virtual machine on top of Windows.

Makes sense
NetTop will run on top of a more secure distribution of Linux that the NSA has developed and an initial version of which it released in December 2000.

While nothing is certain in security, University of New Haven's Professor of Digital Forensics Investigation Fred Cohen said VMware's idea seems to be a good one.

"It makes sense," he said, adding that "the current VMware technology is not up to a level of assurance necessary for this."

To make Linux secure, Cohen said, better handling of various access levels--essentially, the ability to classify data for various secrecy ratings--needs to be added.

But Cohen agreed the decision to run the VMware technology on top of Linux, not Windows, is key to a government agency like the NSA.

In a nod to the open-source community, he said that--for the NSA's purposes--seeing the source code and testing its security is extremely important. "You wouldn't want to do it on Windows NT, because you know nothing about what is going on inside NT," he said.

Also, I wouldn't think of impinging on NSA or HPs interest in a trademark or patent, this would be at a fairly high level.

[astcell]

This may be unrelated, but it reminds me of a discussion I once had with friends in Mensa. Trying to impress one another, discussions started out comparing the recent elections and how they related to 16th century Germany. But by the end of the night we were simply discussing pornography as a religion.

[FerrusGroup]

Mensa?

Well, not really the intent. Want to compare dicks too?

Really, as I said in my original post, I am new here so please forgive this but the content at the last (and my first) DefCon had more to do with aluminum sombreros than security.

To a point it was a good party in the desert but when I read posts about how to get started in the security field, it seems that the community might benefit from some more current topics.

Unless you really really like dodgeball.

[astcell]

FerrusGroup, you may want to link up with jrl. You two have a lot in common. You are both very keen on the industry, you both know your stuff, and you both can lose folks in a blink of an eye. Sort of like algebra in junior high school, you know, how it all goes well then all of a sudden you can be sitting there saying WTF. Forgive us if we are not all on your thinking track. Maybe a little more background for those of us in another field?

[converge]

They caught the last train for the coast
The day the forums died.

[Thorn]

This may be unrelated, but it reminds me of a discussion I once had with friends in Mensa. Trying to impress one another, discussions started out comparing the recent elections and how they related to 16th century Germany. But by the end of the night we were simply discussing pornography as a religion.

ASTCell, your last line indicates pornography isn't a religion. Have I been misinformed?

[allentrace]

They caught the last train for the coast
The day the forums died.

Don Mclean is rolling over in his grave or he would be if he were dead.

[octalpus]

ASTCell, your last line indicates pornography isn't a religion. Have I been misinformed?

I think you misread. He only said that the discussion revolved around pornography as a religion... I would assume it was a discussion of the merits of such.

[TheCotMan]

Has there been any thought to discussing MLS Hybrids, like NSA NetTop? It could be a good entry point for discussing SELinux, local application boundaries (LABs), and the use of virtualization technologies such as VMware and Virtual PC. Also would be good vetting within the community as to sandboxing on the local PC. I can think of friendly vectors for IPSec and L2TP tunneling for enclaves too.

From my perspective, virtualized servers with vmware, et al, does little in the way of added security when each virtualized instance has a root admin that is untrusted, and sharing the same physical CPU (hardware.) True application boundaries with emulated hardware (not virtualization) does have theoretical security advantages, but with added complexity you have increased risk for failure in implementation. In addition, the loss in efficiency in a true emulated hardware system (not virtualization) is sufficient for such systems to not be quickly adopted by the private sector unless security is the highest concern.

As a result, R&D, resources, and development for true emulation systems with true modular separation is smaller due to market saturation by the less secure, but more efficient virtualization as found in vmware and others.

On the topic of COTS replacements, I have issues. While COTS continue to be associated with closed-source, commercial system that are not scrutinized by third parties, I see little advantage in them. Built-in obselecence drives the COTS industry, and while profit is motivated by revision releases and not complete products that are "secure" and free of bugs, it would seem unwise to rely upon them for security.

I do appreciate the view of ThinClient technology and have worked to build such systems at work. They can be easy to support, but the increase complexity decreases reliability; network failure is equitable to station failure.

If you had new information, or a presentation on methods to defeat virtualization instances (beyond DoS) then I think you might have a crowd. If you just plan to review online content and re-package it, you probably won't have a very big audience.

P.S. I have tried to engage members in technical discussions, and learned the hard way that the forums seem to be mostly about 2 things:
1) Arrangement, planning, discussion, events about DefCon (Rides, sleeping arrangements, ideas on events, questions about the convention)
2) Socialization

Occasionally, people will discuss technical items, or rant about something they do not like, but that is uncommon outside of special sections (like wireless.)

Nobody reads my long posts (including this one.) Keeping posts short, is a good way to get people to read them.

One more thing...
I am just a newbie. A lack of response from others does not mean the members here are stupid, or uneducated. There are many people in these forums who have demonstrated to me that they know quite a bit more than me. Read the many archived threads, and you may see what I am writing about. Many members probably just do not have time to read the long posts, and those who do, may not have time to reply.

Good luck.

[d3ad1ysp0rk]

After spending a bit reading all the info you linked to, I have a.. decent understanding of MLS, but I'm still not exactly sure.

Anywho, I'd love to learn more of it, and I'm planning on making the haul to vegas this year. Maine to Vegas, it'll be a trip alright.

[FerrusGroup]

Mitigation Strategies

Thanks for your consideration. If I may make some observations and ask for your thoughts.

From my perspective, virtualized servers with vmware, et al, does little in the way of added security when each virtualized instance has a root admin that is untrusted, and sharing the same physical CPU (hardware.)

I would agree that the NSA is pursuing Class/Unclass implementations out of zeal for the benefits that would be most properly achieved by a true MLS solution and that there is little common work along raising the architectural bar across DoD. Along that line, how would you view current efforts to accomplish such a system?
I have yet to see coherent tagging of legacy data let alone movement off of COTS to anything more open. More to the point, I think the contemporary references to MLS Hybrids are more useful to lesser enclaves, such as SBU vs. Unclass.
I feel like I have to shoot myself now before the interview, so I'll leave that point behind.

The focus of such a talk would be to layout single processor, single core implementations in trivial environments, such as those found in commercial applications. More of aiming for the DefCOn audience I think. It would be fun to review futures, but I think that's past my pay grade.

Built-in obselecence drives the COTS industry, and while profit is motivated by revision releases and not complete products that are "secure" and free of bugs, it would seem unwise to rely upon them for security.

I agree that open source is more subject to review and less subject to unfound bugs than is typical COTS, but that doesn't obviate the current and future reliance on COTS, just it's logical demise if we all join hands and sing. Kumbia my friends, kumbia. Again, that may be off topic ;)

I do appreciate the view of ThinClient technology and have worked to build such systems at work. They can be easy to support, but the increase complexity decreases reliability; network failure is equitable to station failure.

I do agree with the strength and advantages inherent in the TC/SBC realm. But looking at things like SunRay makes me ebrace your point of complexity degrading reliability. I do also believe that without a network, there is only notepad; this in as much as a disconnected information appliance represents major (and unacceptable) failure. For those applications where we would accept SBC solutions, are there really SLAs so low that downtime is an option? I suppose there are always environments that have to operate while embarked but that isn't the point either, nor are those devices the target.

If you had new information, or a presentation on methods to defeat virtualization instances (beyond DoS) then I think you might have a crowd. If you just plan to review online content and re-package it, you probably won't have a very big audience.

Well thats an interesting point. I had assumed that this would be an introductory conversation. If the forum has any life to it maybe the question should be at what level this type of content makes sense. As I said in my original post, I enjoy going aluminum sombreros as much as the... well I don't really, but who is willing to step up? (End of Newbie Enthusiasm) Is there a sustained interest from the group in discussing hardware level (CPU) contamination between processes? That's not so much a virtualization attack as a whole other vector on hardware vulnerability - and thus your point on emulation but I think moreover a wider one on attack surfaces for PC hardware.

Yeah so It's really long, deal with it.

So, what to do with this?

I could always present a class on how VM makes a bitchin classroom environment, but that seems overdone. Is there any interest in a survey of this technology or are we holding out for the fresh content? You can understand my concern when there is such a wide variety of responses to the thread.

I myself think 1:4 monkey butlers is enough.

[TheCotMan]

Along that line, how would you view current efforts to accomplish such a system?

Better than nothing.

I agree that open source is more subject to review and less subject to unfound bugs than is typical COTS, but that doesn't obviate the current and future reliance on COTS, just it's logical demise if we all join hands and sing. Kumbia my friends, kumbia. Again, that may be off topic ;)

This was stated as my reason to avoid mentioning COTS in such a system. For too many people, COTS == commercial, closed source. Instead, I'll say "rely upon an imported layer of abstraction" or choose another phrase. My implied statement was disagreement with use of the word.

looking at things like SunRay makes me ebrace your point of complexity degrading reliability.

SunRay was crap. We test drove it and a few others, but none of them met our needs. We ended up building our own system to support more secure servers and end-user client workstations which mounted (ro) NFS root filesystems over a high-speed intranet, and local RAMDisk for file serving.

I do also believe that without a network, there is only notepad; this in as much as a disconnected information appliance represents major (and unacceptable) failure.

This was not true for our organization, but is true for many (most?) businesses.

For those applications where we would accept SBC solutions, are there really SLAs so low that downtime is an option?

A key business for which uptime and security are both major concern includes financial groups and stock management firms. Downtime of even a few minutes can cost millions of dollars. Significant, publicized security violations can decrease consumer confidence and lead to long term losses. However, you are correct, this is not the norm in the industry. Most places want efficiency and value, while down-time can be scheduled.

Well thats an interesting point. I had assumed that this would be an introductory conversation. If the forum has any life to it maybe the question should be at what level this type of content makes sense. As I said in my original post, I enjoy going aluminum sombreros as much as the... well I don't really, but who is willing to step up? (End of Newbie Enthusiasm) Is there a sustained interest from the group in discussing hardware level (CPU)
contamination between processes?

Perhaps, more general talks will increase population of your audience. If you make your talk too technical, fewer will show up, because most people do not want to admit they don't know something and some just won't be interested; "What's in it for me?" I'll admit my lack of knowledge any chance I get... after all, I am just a newbie. :-)

Yeah so It's really long, deal with it.

I do not fear reading, but you may discourage an audience that could provide you with the feedback you desire, if you keep your posts long. If you are willing and able to accept this cost, then you have a market with fewer consumers of your content.

Is there any interest in a survey of this technology or are we holding out for the fresh content? You can understand my concern when there is such a wide variety of responses to the thread.

If you had a presentation that went beyond the published content, I would likely attend, but you should know that I am probably a minority in this.

Here is what I would do if I were you... restate your presentation in an "executive summary" which is 3 setences long and use simple language; you may get more feedback from the "old timers" and organizers.

P.S. Don't let the seemingly silly subjects cause you to assume those who post about "monkey butlers" are lame; for some people, these forums are an escape from their professional work. They read enough for education while working that they may choose "fun" over "not fun" and things that remind them of work may not be fun.

Last item: taunting people to react can get you labeled as a troll, and nothing good will become of that. :-/ Understanding how far you can go with taunting people and have it accepted as harmless fun seems to come with how long you have been on the forums and how well people know you.

[murakami]

Bah, I would like to see the architecture of a "true" MLS.

Screw implementations. NetTop has been kicking around forever but its still pretty much prototype, which should tell you that it doesn't meet the requirements.

Airgap still rules.

[Adrenaline]

Major League Soccer kicks ass!
oh. wait, wrong MLS.
:)