Linked by Thom Holwerda on Wed 17th Oct 2012 23:48 UTC, submitted by poundsmack
Privacy, Security, Encryption Kaspersky is working on its own secure operating system for highly specialised tasks. "We're developing a secure operating system for protecting key information systems (industrial control systems) used in industry/infrastructure. Quite a few rumors about this project have appeared already on the Internet, so I guess it's time to lift the curtain (a little) on our secret project and let you know (a bit) about what's really going on." More here.
Order by: Score:
Comment by quackalist
by quackalist on Thu 18th Oct 2012 00:41 UTC
Member since:

Was reading the blog earlier and though it seemed to make sense.....who wouldn't want critical systems to be secure though I did think his " And then there are some details that will remain for certain customers’ eyes only forever, to ward off cyber-terrorist abuses." somewhat contradictory as the clearest example of "cyber-terrorist abuses" have come from some of those selfsame "customers" and unless the OS is secure from their eyes it might as well not exist. Anyway, if its not secure unless by 'obscurity', if you can't trust those in the know not to leak or use that info nefariously, than can it be secure?

Not claiming any great knowledge on how to secure OS's, not at all, it just seems not quite right.

Edited 2012-10-18 00:48 UTC

Reply Score: 3

RE: Comment by quackalist
by Doc Pain on Thu 18th Oct 2012 05:18 UTC in reply to "Comment by quackalist"
Doc Pain Member since:

Anyway, if its not secure unless by 'obscurity', if you can't trust those in the know not to leak or use that info nefariously, than can it be secure?

If you want to be scared to death, visit your local hospital:

Medical devices are a domain for closed-source software. That software may be essential to life of people. So if you are a "cyber-terrorist" and want to hurt "ordinary people", you could take down hospital devices. Everything you need is in there: proprietary devices, often sloppily engineered (from the software aspect), insecure and exploitable; IT infrastructures happily carrying out your orders (PCs, printers, networking gear); people - some stupid, some ignorant, some knowing, but with a voice to "unimportant" to make any change to the status quo, and those in charge of "decision & responsibility", relying on outsourcing, cheap renting, and delegating the own security to 3rd parties who have no other interest than eating from the cake of money, by not really delivering good services. It's not even hard: Bring a prepared USB stick, put it in some unsecured PC, or deal with the WLAN. There's enough old and old-fasioned hardware and software still in use, considered "not that bad", so nothing is questioned, because it "just works". There can't be security without knowledge, and knowledge is usually "left to others" who, in the end, don't really care. And it's not just about the danger of "cyber-terrorism"; just think what you could earn by obtaining patients' and employees' data (personal data, payment details, medical records, pricing, contracts with 3rd party services, data from research studies etc.) and selling them to spammers, advertisers or competitors.

Why can I make those claims? Because I've seen it. Here in Germany. Too often.

Reply Score: 4

Comment by shmerl
by shmerl on Thu 18th Oct 2012 01:22 UTC
Member since:

I didn't quite understand whether it's going to be open source or closed?

Reply Score: 2

Not exactly new?
by Gullible Jones on Thu 18th Oct 2012 01:33 UTC
Gullible Jones
Member since:

The PikeOS microkernel is apparently designed for similar stuff, and is already formally verified.

Not that competition in the uber-secure embedded OS market would be in any way a bad thing, mind...

Reply Score: 3

RE: Not exactly new?
by fithisux on Thu 18th Oct 2012 09:30 UTC in reply to "Not exactly new?"
fithisux Member since:

The rise of microkernels. I believe contrary to Linus that uKernels and user space drivers are a good thing. It needs proper design though and support from hardware and devices.

Reply Score: 5

RE[2]: Not exactly new?
by Gullible Jones on Thu 18th Oct 2012 11:32 UTC in reply to "RE: Not exactly new?"
Gullible Jones Member since:

Different roles IMO. Linux is a whopping big kernel with tons of features - good for servers and desktops, bad for realtime stuff or heavy-duty embedded use.

Linux is the sort of OS you run your server on. PikeOS is the sort of OS you run your Linux on.

Reply Score: 2

Comment by marcp
by marcp on Thu 18th Oct 2012 08:59 UTC
Member since:

On the marketing and financial side - great move. This will might give them monopoly and tons of money.

Now, to the core of the problem:
- looks like "security through obscurity" to me, to some degree. They won't share the code, and they'll have limited number of eyeballs looking at their code [mostly internal contractors]. Good luck with that, Kaspersky
- not based on an axisting code? wow, now that's huge. I don't think they realize the scope of this problem. They'll probobly get some BSD-licensed code anyway [TCP/IP, etc, although this particular one might not be needed]. Reinventing the wheel isn't the most efficient and best way to create anything. Besides - they'll introduce tons of bugs, and they're gonna be on their own with fixing it. We're gonna hear some freaking hilarious news from that front. Mark my words
- no mistakes in kernel code? oh, come on ... it shows you have no idea what you're talking about, Eugene. You CAN'T avoid mistakes as long as the code is being written by humans. And there's no other way to produce code. The code was 'invented' by humans. Monkeys cannot code. Computers could be coding, but they are ... coded by humans. Period.
- minimum amount of code in kernel - reasonable assumption. However, this will not protect you from some ugly zero-day flying around unnoticed. No matter how much code do you got there - it may be always a very bad piece of code [even if you don't know it yet]
- "In such an environment there needs to be a powerful and reliable system of protection that supports different models of security. " - DETAILS, please. This is marketing crap.

I think this is going to be a huge failure. Some people will desperately want to lay their hands on this so it can be broken, mangled and used against companies which use it. Come on - these are the critical systems - power plants, water pumping, etc. Do you expect cyber mercenaries and all of the other baddies to stay away from this?

Good thing is that something finally changes. Running SCADA on top of the Windows OS was like ... ok, I don't know, don't get me started on this, but it was just plain stupid.

Reply Score: 2

RE: Comment by marcp
by lucas_maximus on Thu 18th Oct 2012 09:53 UTC in reply to "Comment by marcp"
lucas_maximus Member since:

- looks like "security through obscurity" to me, to some degree. They won't share the code, and they'll have limited number of eyeballs looking at their code [mostly internal contractors]. Good luck with that, Kaspersky

Can we stop with the persistence of the many eyes principle, please.

More people looking at the code doesn't help you if they don't know what they are looking for.

Reply Score: 3

RE: Comment by marcp
by Gullible Jones on Thu 18th Oct 2012 11:38 UTC in reply to "Comment by marcp"
Gullible Jones Member since:

Money yes, monopoly no. Like I said above, I don't think this is a new idea (though a new implementation certainly wouldn't hurt).

Re FOSS, I honestly don't think open/closed source models make a difference; somebody will eventually put your code under a black-box debugger no matter what.

Reply Score: 2

by jefro on Thu 18th Oct 2012 19:32 UTC
Member since:

The country of hackers is making a secure system?

Reply Score: 1

This is likely going nowhere
by coreyography on Fri 19th Oct 2012 02:08 UTC
Member since:

Industrial control systems used to use purpose-built OSes at all levels, primarily to achieve the realtime performance required. (Some of these still exist, like VxWorks and QNX -- well, I assume QNX, if RIM hasn't made it into a smartphone toy.) A bit later, they moved the console/server layer to VMS or commercial Unix (Solaris, HP/UX). Now, almost all such systems have devolved into running some Windows variant at the server/console layer, despite the frequent objections of technically-minded users. The vendors could enjoy wide compatibility, and cut their costs by relying on prebuilt tools and libraries for Windows (though Microsoft's API-du-jour mentality burned them more than once). And of course Windows had a lot of CEA - clueless executive appeal.

Even though malware has made Windows a much bigger liability, I don't see the ICS vendors going back. There's no widely used, commercially-supported alternative OS, and the vendors are happy to sell you add-ons (virus scanners, whitelisting software, firewalls, and $ecurity $ervice$) to protect your control system's soft underbelly. Further, you can do a reasonably good job of protecting your system through good security practices and procedures; most of the ICS "hacks" are a result of weaknesses in these practices and procedures. If that's not enough, there are even pricey, physically-enforced one-way firewalls for certain segregation requirements (e.g., NERC). And contrary to what Kaspersky says, you _can_ run these systems isolated, or at least on their own private (control) network. Plants I worked in did it for years; all this connectivity is a relatively recent phenomenon.

At the controller level, the proprietary, realtime OSes are still used. I don't think Kaspersky even tries to address the realtime requirements; I didn't see it mentioned in their article.

Reply Score: 1

by poundsmack on Fri 19th Oct 2012 22:45 UTC
Member since:

They could have used QNX at the core and saved themselves a lot of time.

...but then again, if I had my way QNX would power everything. ;)

A man can dream can't he? A man can dream...

Reply Score: 2