Hacker News Reader

Comment by brettbender | Hacker News Reader

At least now the next time a client wants me to do something in .NET I have a good excuse to gently persuade them to something else (until this gets patched, at least).

View on HN

Then you should gently persuade them away from JavaServer Faces and Rails too. The same vulnerability exists in both.

No, Rails encrypts then signs, which means it verifies HMAC before it CBC-decrypts the message. It isn't vulnerable to this attack. What led you to say that it was?

JSF was vulnerable to this attack; Thai and Juliano, the researchers that other HN users are trashing on this thread, also found that flaw as well. I have to assume it's been fixed by now.

What led you to say that it was?

Section 5.1 of this whitepaper[1] by Rizzo and Duong, that lists Ruby on Rails as one of the web frameworks vulnerable to this attack.

Granted, this paper was published in May, and I'm not a Rails guy, so it's quite possible this has been patched as well.

[1]http://usenix.org/events/woot10/tech/full_papers/Rizzo.pdf

Their comment about MessageEncryptor is accurate, but I don't believe the bare #encrypt (which could create a vulnerability if exceptions weren't laundered) is used anywhere. Rails sessions are encrypted and then HMAC'd. This pattern isn't susceptable to padding side channels.

This exploit has remarkably little applicability to the vast majority of ASP.NET applications. These guys are grossly overblowing this to try to get attention.

The default "user is logged in" cookie in ASP.NET is vulnerable to this, since that cookie value is just an encrypted username. An evildoer (even a relatively unskilled one) could use this attack to login to an ASP.NET app as anyone they wanted.

To decrypt it you need an exception message...so turn on custom errors and you're protected it seems. Assuming you don't do that, once you decrypt it you get what, a session ID? Then what?

You apparently don't need the exception message at all. This was all discussed on the thread 3 days ago: http://news.ycombinator.com/item?id=1687547

Besides what storm said: decrypting isn't the problem, being able to encrypt a new cookie is. If I can "bradhe" as my username in my own cookie, the default ASP.NET forms auth will consider me logged in as you.

All they need is distinguishable errors when decrypts fail. The contents of the error are irrelevant. You can send a bare 500 with no content and still be trivially vulnerable to this attack.

Fair enough, I guess I didn't understand the attack thoroughly.

Exactly how do you know that? Juliano used his pre-existing tool (POET) to get admin privileges on DotNetNuke, without ever having used DNN --- or, presumably, modifying his tool. DNN is not an obscure .NET app.

What makes you make a comment like this? What evidence are you basing it on?

>Exactly how do you know that?

I have never worked on an ASP.NET-based site that relied upon cookies, encrypted or not, for anything. Even where the built-in forms authentication was used the username always had correlating server-side state that was triumphant.

So saying that you can modify cookies == a complete and utter non-issue for any site that followed any reasonable security practices.

The easy answer here is to say, "you mean besides DotNetNuke".

The deeper answer would be to point out that my day job is security evaluation for apps, roughly 50% of which are Fortune-100 web applications, roughly 60% of which are .NET applications, and you're flat-out wrong. There's a reason why cookie/session security is #3 on the OWASP top 10.

You're also presuming that the issue being discussed is, specifically, a cookie.

I understand why you're as defensive as you are; you're a professional .NET developer and Hacker News is hostile to Microsoft and, especially, .NET. Listen to me: I am not hostile to Microsoft or .NET. Microsoft is a client of ours. They do software security better than any company in the industry. Take my word for it: it appears that they screwed this one up just like everyone else that tries to encrypt with AES.

>The easy answer here is to say, "you mean besides DotNetNuke".

Yes, I see you've already mentioned DNN. It's good that you have that example. It would be interesting if I somehow claimed that every app is immune to it.

Countless apps have countless vulnerabilities.

>You're also presuming that the issue being discussed is, specifically, a cookie.

I know that the issue being discussed is specifically a cookie. It's specifically an AES-encrypted cookie.

To your edit: Defensive? Hardly. I'm just bitter after an endless stream of B.S. security claims by the security industry. It always follows the same pattern of arm waving and press releases, with a promised demonstration, and then the day of reckoning comes and...quiet. Maybe this will be the exception but, we'll see.

This has nothing to do with any sort of loyalty to .NET, though I hardly find HN to be anti-.NET, or anti-Microsoft for that matter. Whatever, in any case. It isn't my fight to wage.

You have their deck or their paper? URL? I'm a casual acquaintance of both Juliano and Thai, and friends with other Netifera people, and so all I have to go on is what they've told me. I'd be interested in reading the actual material.

You have it, right? You sound like you do. Otherwise, how would you know how serious the issue is?

That's a nice shtick (the defensive bit was my favourite addition). It's also facile.

They specifically detailed that it's an AES cookie encryption attack, yet you're acting like I'm going on a limb saying that?

Where did they specifically detail that it is an AES cookie encryption attack?

They didn't; this commenter has decided that he can easily infer the details of a presentation he hasn't seen from how it's reported in the trade press.

So it isn't exactly an AES Cookie Encryption Attack, it is actually worse: http://www.ekoparty.org/juliano-rizzo-2010.php

The most significant new discovery is an universal Padding Oracle affecting every ASP.NET web application. In short, you can decrypt cookies, view states, form authentication tickets, membership password, user data, and anything else encrypted using the framework's API!

"It always follows the same pattern of arm waving and press releases, with a promised demonstration, and then the day of reckoning comes and...quiet."

A few minutes ago, live on stage at a security conference in Buenos Aires (#ekoparty) they popped local SYSTEM privileges remotely on both DotNetNuke and SharePoint installed in a typical production configuration.

In case you think they stacked the deck, those applications were chosen only a couple of days ago after Juliano asked on Twitter for suggestions for the presentation.

What do you mean by "local SYSTEM"?

I eagerly look forward to details on it. Where might I find them?

> What do you mean by "local SYSTEM"?

The highest system privilege level on Windows. They were able to interactively run CMD.EXE as the LocalSystem account on the remote web server.

> I eagerly look forward to details on it. Where might I find them?

The details were not disclosed until today when the attack was presented at a security conference. As far as I know there isn't anything available online yet, but that should change very soon.

I think you owe Juliano and Thai an apology. Observers can read downthread to see why.

You have their deck or their paper? URL? I'd be interested in reading the actual material.

You have it, right? You sound like you do.

I wonder how I might have any details about this. I mean, I'm just some random guy on the Internet just like you. Surely we must come to our convictions about things, like CBC padding attacks or web frameworks or the best internal temperature to cook meat to, the exact same way.

I think it's classy to call out two people you don't know as grandstanding opportunists based on zero knowledge of what it was they were doing and then argue with the guy who calls you out on it. I'm weird that way. I'm going to get downmodded for being a snarky asshole here, but they'll be negative karma points well worth it for my sanity.

Technically you change the cookie encryption method to 3DES in the server's config file to work around it. But, I would try to "gently" persuade anyway.

I can't understand why 4 people voted this comment up. The problem is with the CBC block cipher mode; it's not with AES. Who's writing recommendations that say "switching ciphers fixes the problem"?

Who's writing recommendations that say "switching ciphers fixes the problem"?

Visual Studio Magazine [1].

I've enjoyed reading all your comments on this thread. You clearly know your stuff. Based upon what you've seen so far and what you know about the ASP.NET framework, do you see any way around this short of some patch from Microsoft?

[1] - http://visualstudiomagazine.com/articles/2010/09/14/aspnet-s...

The funny part about this story is that Thai's video (of him breaking DotNetNuke) breaks a 3DES key. The great thing about CBC padding and bitflips is that it's inherent to the block cipher mode; switch to DES-EDE, MARS, Twofish, or Serpent and you still have the same vulnerability. You don't even need to know how the algorithm is implemented! The same exploit will work, regardless of the underlying exploit.

You gotta love crypto.

Underlying cipher, sorry.

Reply on news.ycombinator.com