[Copied from anti-fraud]

| 2.1 Current situation
|
| Currently this technology seems to be only in the hands of financial
| fraudsters. It is currently high-tech, and high-priced, with very
| limited distribution. The time-to-market for active attacks is
| estimated to be 3 days.
|
| Only targets of significant financial value have a worrying risk of
| being attacked with this technology in the near-term.

I think you got that backwards. Currently, this technology is only required to attack customers of some financial institutions. For the majority of web applications, it is sufficient to eavesdrop on HTTPS connections and recover the protected storage area, and you can issue arbitrary transactions.

| 3.1.4 Class 4 SmartCard reader
|
| A theoretical solution for a secure client would be the Class 4
| SmartCard reader. Although not available, the device characteristics
| and security profile are relatively well understood, and appropriate.

FINREAD does exist, but failed to penetrate the market. There are also relatively cheap gadgets which use symmetric cryptography and require that the user transfers a couple of bits manually to and from the device (which means that there is zero setup cost on the customer side).

[Copied from anti-fraud]

Section 1.2.1:

A "plugin" is not the same as a Browser Helper Object. A plugin is something like Flash or Java, which can take over a square shape in your web page and display its own content.
http://en.wikipedia.org/wiki/Plugin

What you are talking about are "addons" or "extensions".

3.1.1 is a dead loss. Even if you could create it:

a) Users would never use it
b) Fraudsters would just intercept and change the HTML at the network layer, or replace the binary entirely.

3.1.3: how does this help? Running your OS inside another OS does not, in itself, provide any protection from you getting a trojan installed inside the guest OS.

And an idea you've missed out (as far as I can see), which some banks are already using - a challenge/response system where the browser prints up a challenge number, you type it into the keypad on a smartcard reader with your bank card inserted, it does some crypto and offers a response, and you put that back into a text box on the computer. In other words, you move the trusted platform off the PC. Bad usability, but reasonable security.

"Once an attacker has root access on your computer, it's not your computer any more."

Re Virtual machines (3.1.3) -- while it is true that no guest VM can be more safe than the host, the typical way to use them is not as a secure isolated system, but as a known-to-be-unsecure one from which the host system is isolated. Example: you copy a pristine VM, surf to the unsafe website, and once you are done with whatever it is you wanted to do, discard the now tainted copy.

Hi Philipp,

I will probably have other comments to make, but here are a couple about your mention of PassFaces -- the description in Section 4.3.3 misses a few worthwhile points:

- It is hard to fool a user into revealing their secret just by asking for it (e.g. "please enter your password"), because the PassFaces prompt is an array of nine faces, exactly one of which must be the correct face.

- It is defeated by a man-in-the-middle attack (the attacker can obtain each array of nine faces by attempting a login at the real site).

- Although it seems significantly better than typical passwords in terms of memorability/strength trade-off for a single site, it may be an excessive burden for users to have secrets made of PassFaces at many different sites.

Hope these are helpful,

Philipp Gühring wrote:

>> 2.1 Current situation
>>
>> Currently this technology seems to be only in the hands of financial
>> fraudsters. It is currently high-tech, and high-priced, with very
>> limited distribution. The time-to-market for active attacks is
>> estimated to be 3 days.

Like Florian, i am skeptical of this claim. I have seen screenshots of downloadable tools that will generate custom Trojans for you -- you choose the options you want, and they will produce a ready-made Trojan. Hardly any expertise is required.

more complications in developing a secure browser:

http://www.theinquirer.net/default.aspx?article=32756

has a good story about firefox scripting engine, and mozilla seamonkey and so forth ...

Under 3.1.2 Unwritable Distribution, the author seems to have missed out mentioning write-protected USB pen drives on which the browser is installed and from where it runs. So the user wouldn't need to restart from a bootable CD...

Hmmm, of course the reg keys, app data and user settings would be on the hard drive and could be tampered with.. Option could be to use this as an alternate browser for critical sites, while doing daily browsing from one installed on the hard drive as usual.

>>A "plugin" is not the same as a Browser Helper Object. A plugin is something like Flash or Java, which can take over a square shape in your web page and display its own content.
http://en.wikipedia.org/wiki/Plugin
-----------

Any solution that relies on browser content pane (rectangle) will not work. Man in the middle can manipulate this. This approach might work against man in the browser. KeyID (http://www.KeyID.com) has recently lauched a product that works by integrating authentication and communication channel. The solution relies on a browser plug-in where you type in the username, password, OTP etc. The authentication information is not typed in the browser content pane. The authentication information in then mixed with the channel information and submitted on a standard https connection. It uses same form fields and does not require any significant changes on the server end.

The same product also works with Smartcards using the same concepts.