Testing Commercial 2-Factor Authentication Systems?

Fry-kun asks: "I recently became interested in setting up a 2-factor authentication system for my laptop. With that in mind, I bought a fairly inexpensive USB key. Although it seems to work, I can't bring myself to trust it completely: Kensington claims that the system is secure, but there is no independent security lab analysis of the product. In other words, for all I know, there may be a gaping hole in their security setup. Worse yet, there are apparently no reviews of the product, no mention of anyone trying to test it and no hardware hackers tried to make it work in Linux, even though it's been out for over 2 years. How would you go about making sure that a security product does what it claims to?"

Comments

Testing commercial security

[mandelbr0t]

You can't. All security software needs to be OSS for this reason.

That being said, OSS had a 2-factor authentication mechanism available years ago. Encrypt your hard drive, save the key to a USB key and enter a passphrase. You'll need to both insert the USB key and type your passphrase for the root disk to get mounted. That's pretty much the entire system locked down.

This article appears to detail that process.

You don't

[Anonymous Coward]

I am posting this as AC because I do this for a living for a large government agency.

You are not sure, which is the problem. I will give a nod to Kensington here, though. They are about to make a lot of money because they are serious about security, unlike a lot of other companies that peddle USB devices (Kangaroo, I am looking your way).

While it is commendable you are looking for two-factor authentication, a USB key is not the way to go here. The goal here is to not be able to break your encryption if you are forced or influenced to give up your password. Any system you can set up yourself will be breakable by you unless you take extreme measures. For the sake of argument, we will assume that there are no extreme measures in place, but your encryption can still be cracked by you.

Your best bet here is to go with full disk encryption. For further security, use truecrypt with a file on a CD or USB device as part of the key, as was referenced above.

For further security, encrypt again.

As you can see, this goes on. The weak point is you. If you can break it, you can be forced to break it.

If you want complete deniability, triple encrypt all of you regular data, then quadruple encrypt your sensitive data somewhere else. Use files, passwords, obfuscation, etc.

You will still be better off than most people. Including the government, according to plenty of stolen laptop press reports.

Re:RTFQ

[MarcoAtWork]

having a physical USB token with a TC volume (esp. the kind that stores things in a steganographic way) is in my opinion practically equivalent security-wise to the article's 2-factor authentication if you're smart enough to have your token on your keychain or something (a lot more likely than somebody will steal your laptop than your token IMHO).

In any case if you want to increase the security of what I proposed nothing forbids you from getting TWO usb tokens, create truecrypt volumes on both of them, and then create an overlaid raid-0 striped partition on both of them: in this case an attacker would need to steal BOTH tokens and BOTH passwords to gain access to your files.

Schemes like these make it also very easy to mandatorily have multiple people there to open the files (say, all the directors, etc.). If you do things like RAID-5 you could also make it so that you could still access the information with N-1 USB tokens (in case one is lost).

I do think that these solutions are safer than trusting a random crypto vendor, also this is why I have all my sensitive things (tax returns etc.) strictly on TC volumes.

You can't evaluate it, and it probably sucks

[swillden]

I work as a secure systems designer and consultant, and I've had some opportunities to review the security of commercial systems of various sorts. What I've learned is (1) properly evaluating commercial security tools is nearly impossible and (2) much of it is lousy.

The most effective means I've found of evaluating tools is to have a client sitting on a really huge purchase order, so that the vendor will give me access to key security personnel on their design, development and testing teams in order to make the sale. The people in question won't actually answer my detailed questions, in most cases, but I can still get a feel for how they think, and what they consider important. That actually gives me a pretty good idea of how secure the stuff they build is, though it's not as good as actually doing a detailed analysis of the design and implementation. Ideally, I'd like to talk to their people, do a detailed analysis of their designs, perform a cursory review of their implementation and then really, deeply scrutinize their security design and QA processes.

What I've found when I start pushing to talk to the "security guys" is that in surprisingly many cases there are none! Or there was one, but he left. Or there is one everyone thinks is the security guy, but he's really just a developer with a basic understanding of security principles, no time to really focus on security, and no authority to get any security problems he finds fixed.

Note that this is not always true. I've found some companies that do a really good job, but they're definitely in the minority.

Assuming you can't actually force the vendor to let you talk to their security team, the only thing I can suggest is that you start looking at publicly-available information. Some things to look for are:

1. Do they have any serious, well-regarded security researchers with solid publication histories? If so, then you know that at least someone at the company has a clue. Then if you can determine whether or not the clueful people are allowed anywhere near the product you're interested in, you may learn something useful.
2. How open are they about their product designs, especially the security features? The more detailed technical information they provide, the better. The more they hide behind secrecy and buzzword bingo, the more you want to steer clear.
3. What kind of a company is it? I like companies whose whole focus is security and are relatively large (for security companies) and well-established. Big companies that can (and do) have large security-focused groups are good too.
4. Have there been any published analyses of the products? I'd rather use a product that has had significant scrutiny and a few security defects identified (and fixed), than one that no one has ever bothered to look at. This actually goes back to the previous point. Products from established companies with a focus on security tend to get scrutiny.