Onderwerp: Re: [Full-Disclosure] Clear text password exposure in Datakey's tokens and smartcards

=2D----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello,

In the wise words of vuln@hexview.com, on Wednesday 04 August 2004 07:08:
> Clear text password exposure in Datakey's tokens and smartcards
[SNIP]

> Cause and Effect:
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3 D
> The communication channel between the token and the driver is not
> encrypted. User's PIN can be retrieved using proxy driver or hardware
> sniffer.
Note that this is true for almost all card readers on the market, not only =
for=20
Datakey's. Having worked for companies using crypto smart cards, I have=20
conducted a few risk analysis about that. The conclusion has always been th=
at=20
if the PIN must be entered from a PC, and the attacker has means to install=
=20
software on the system (through directed viruses, social engineering, etc),=
=20
the game's over.

The only solution against that problem is to have the PIN entered using a=20
keypad on the reader. Only then does the cost of an attack raise=20
significantly. But that is opening another can of worms, because there is=20
(was?) no standard for card readers with attached pin pad (at the time,=20
PC/SCv2 wasn't finalised - is it?).

[SNIP]

Cheers,

Lionel

=2D --=20
"They that can give up essential liberty to obtain a little temporary
safety deserve neither liberty nor safety." -- Benjamin Franklin

Lionel Ferette
BELNET CERT Coordinator

Rue de la Science 4 Tel: +32 2 7903385
1000 Brussels Fax: +32 2 7903375
Belgium PGP Key Id: 0x5662FD4B
=2D----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFBEIYGDd3gqVZi/UsRAqEMAKDAISNaTuvH8eH37ER1wSO/zUq22gCgsG9W
PqY79HOMC3f+CWkUQXLPp1E=3D
=3Dk9PO
=2D----END PGP SIGNATURE-----

Surely if the user is entering a passphrase then the same problem exists -
that of effectively eavesdropping that communication from the keyboard?

Ignoring the initial expense for a moment, wouldn't it have made a lot of
sense to include the keypad actually on the cards? Obviously, card
readers would need to be contructed such that the keypad part of the card
would be exposed during use. The keypad security could then rely on the
tamper resistant properties of the rest of the card.

From a costs perspective, I would guess that the actual per-card cost
increase would be minimal if hundreds of millions of these cards were
produced.

Kev


> Lionel Ferette wrote:
>
>> Note that this is true for almost all card readers on the market, not
>> only for Datakey's. Having worked for companies using crypto smart
>> cards, I have conducted a few risk analysis about that. The conclusion
>> has always been that if the PIN must be entered from a PC, and the
>> attacker has means to install software on the system (through directed
>> viruses, social engineering, etc), the game's over.
>> The only solution against that problem is to have the PIN entered
>> using a keypad on the reader. Only then does the cost of an attack
>> raise significantly. But that is opening another can of worms, because
>> there is (was?) no standard for card readers with attached pin pad (at
>> the time, PC/SCv2 wasn't finalised - is it?).
>>
>
> at least some cards are supporting des passphrases to implement secured
> communication channels but I suppose this feature is not that widely in
> use.... how many card owners are prepared to remember both PIN codes
> and passphrases...
>
> toomas
>
>



--
Kevin Sheldrake MEng MIEE CEng CISSP
Electric Cat (Bournemouth) Ltd

I don't doubt the widespread inherent nature of this vulnerability.

Smart cards used with PCs as the man-machine-interface is simply asking
for trouble IMHO. Unless I'm mistaken, there is very little that can be
done about this without redesigning/replacing current smart cards.
Obviously, bespoke MMIs (such as door access systems) are less vulnerable
to this issue. My problem is that people seem to view smart cards as a
panacea, as a place to store digital certificates or 'identities' and then
use them to 'prove' who they are.

Don't even start me on biometrics; if there is a PC (or anything that runs
software) between the reader and the authenticator then couldn't you just
replace the reader with another bit of software that pretends to be a
reader? Where can I get the biometric info required? Police stations
have a large quantity of finger prints, as have practically every surface
you've touched recently...

Kev



> This exposure, of PIN compromise, is genric in all smartcard products
> today,
> unless a dedicated PINpad or biometric-sensor equipped readers are used
> -
> putting cost of ownership towards $1000 in some cases.
> PC/SC doesn't help - as a data interfcae API spec, it excludes human
> interface aspects. STIP (Small Terminal Interoperability Platform at
> www.stip.org) moves in this direction, but has evolved into many
> variants to
> interoperate with proprietary vendors and proprietary industry standards.
>
> The challenges in putting biometric sensors or PINpads onto cards include
> the need to conform to ISO 7816 for form factor, physical resilience etc,
> and that the cards are unpowered. Or, someone redesigns the entire
> form-factor, user interface model, portability and business model -
> something that has previously failed to go anywhere.
>
> Something like a mobile phone or PDA is a good compromise tool to this
> overall exposure, imho.
>
> Lyal
>
>
>
> -----Original Message-----
> From: Kevin Sheldrake [mailto:kev@electriccat.co.uk]
> Sent: Thursday, 5 August 2004 8:39 PM
> To: Toomas Soome; lionel.ferette@belnet.be
> Cc: vuln@hexview.com; full-disclosure@lists.netsys.com;
> bugtraq@securityfocus.com
> Subject: Re: [Full-Disclosure] Clear text password exposure in Datakey's
> tokens and smartcards
>
>
> Surely if the user is entering a passphrase then the same problem exists
> -
> that of effectively eavesdropping that communication from the keyboard?
>
> Ignoring the initial expense for a moment, wouldn't it have made a lot of
> sense to include the keypad actually on the cards? Obviously, card
> readers would need to be contructed such that the keypad part of the card
> would be exposed during use. The keypad security could then rely on the
> tamper resistant properties of the rest of the card.
>
> From a costs perspective, I would guess that the actual per-card cost
> increase would be minimal if hundreds of millions of these cards were
> produced.
>
> Kev
>
>
>> Lionel Ferette wrote:
>>
>>> Note that this is true for almost all card readers on the market, not
>>> only for Datakey's. Having worked for companies using crypto smart
>>> cards, I have conducted a few risk analysis about that. The conclusion
>>> has always been that if the PIN must be entered from a PC, and the
>>> attacker has means to install software on the system (through directed
>>> viruses, social engineering, etc), the game's over.
>>> The only solution against that problem is to have the PIN entered
>>> using a keypad on the reader. Only then does the cost of an attack
>>> raise significantly. But that is opening another can of worms, because
>>> there is (was?) no standard for card readers with attached pin pad (at
>>> the time, PC/SCv2 wasn't finalised - is it?).
>>>
>>
>> at least some cards are supporting des passphrases to implement secured
>> communication channels but I suppose this feature is not that widely in
>> use.... how many card owners are prepared to remember both PIN codes
>> and passphrases...
>>
>> toomas
>>
>>
>
>
>



--
Kevin Sheldrake MEng MIEE CEng CISSP
Electric Cat (Bournemouth) Ltd

Not unless the card is stolen and the owner either doesn't notice
immediately or doesn't report it immediately. How many people will turn
up at work (for instance) claiming to have 'forgotton' their card rather
than report it lost, on the off chance they have actually misplaced it?
If the keys give access to money, reputation, authority or the like then
perhaps the size of the exposure window is important?

Kev


> Perhaps I'm missing something here. As far as I can tell, no keys
> located on the card were compromised, only the PIN was. Since this is a
> two factor authentication system, possession of the PIN is of little
> value without possession of the token itself.
>
> Am I missing the point here?
>
> regards,
>
> -lee
>



--
Kevin Sheldrake MEng MIEE CEng CISSP
Electric Cat (Bournemouth) Ltd

"Kevin Sheldrake" <kev@electriccat.co.uk> wrote:

> Ignoring the initial expense for a moment, wouldn't it have made a
> lot of sense to include the keypad actually on the cards?
> Obviously, card readers would need to be contructed such that the
> keypad part of the card would be exposed during use.

No, they wouldn't. The card could remember the key typed on it for,
say, 60 seconds.

Seth