WEBVTT

00:00:00.670 --> 00:00:02.346
I'm conscious we can get really boring on crypto.

00:00:02.346 --> 00:00:03.759
I really want to do one more thing on crypto.

00:00:03.759 --> 00:00:04.516
We should.

00:00:04.516 --> 00:00:06.808
Right, which is this; we always ask the question,

00:00:06.808 --> 00:00:11.037
is this a way that people get compromised?

00:00:11.037 --> 00:00:19.353
You know, I think this is the more complex attack vector.

00:00:19.353 --> 00:00:24.287
I think this is where you're talking about risks and rewards, right?

00:00:24.287 --> 00:00:27.058
The average, you know, burger joint, laundromat,

00:00:27.058 --> 00:00:31.606
hotel, the attacker is going to go after the data in transmission,

00:00:31.606 --> 00:00:32.835
right, which we'll talk about next.

00:00:32.835 --> 00:00:33.312
Now.

00:00:33.312 --> 00:00:34.624
Right.

00:00:34.624 --> 00:00:35.217
That's really interesting.

00:00:35.217 --> 00:00:39.182
So when I was in the card brands, the attack was against stored cardholder data,

00:00:39.182 --> 00:00:44.098
because people did store cardholder data then in ways that they don't now,

00:00:44.098 --> 00:00:45.997
and they were always going, oh, it's encrypted,

00:00:45.997 --> 00:00:47.848
and the encryption keys were always taken because they

00:00:47.848 --> 00:00:49.204
were still in the same database.

00:00:49.204 --> 00:00:49.911
Right.

00:00:49.911 --> 00:00:53.684
So three or four years ago this was a massive cause of compromise.

00:00:53.684 --> 00:00:56.453
The people would take SQL databases with the keys in a

00:00:56.453 --> 00:00:58.751
text file mixed in the SQL database, or even better,

00:00:58.751 --> 00:01:02.250
in the SQL database in a table called Encryption Keys.

00:01:02.250 --> 00:01:03.029
Right.

00:01:03.029 --> 00:01:09.829
Now I would say the sort of primary ways of handling

00:01:09.829 --> 00:01:14.864
payment card data and encrypting it, that attack vector is less easy.

00:01:14.864 --> 00:01:21.362
There are still organizations that, and this frosts me to no end,

00:01:21.362 --> 00:01:21.576
right,

00:01:21.576 --> 00:01:27.910
that lean on the disk encryption or volume encryption subset of

00:01:27.910 --> 00:01:32.231
PCI DSS for servers that are kind of always on,

00:01:32.231 --> 00:01:32.525
right?

00:01:32.525 --> 00:01:34.525
So just to back up.

00:01:34.525 --> 00:01:36.115
This is requirement 3.4?

00:01:36.115 --> 00:01:37.439
Maybe 3.1.

00:01:37.439 --> 00:01:38.421
Let's just get it on the screen.

00:01:38.421 --> 00:01:39.841
Yeah, let's do that.

00:01:39.841 --> 00:01:42.496
This is one that absolutely drives me bonkers.

00:01:42.496 --> 00:01:43.375
The problem is,

00:01:43.375 --> 00:01:47.196
so my view of this requirement is it doesn't know what it's trying to do.

00:01:47.196 --> 00:01:50.484
So if you encrypt a hard disk, right?

00:01:50.484 --> 00:01:52.294
My laptop hard disk is encrypted.

00:01:52.294 --> 00:01:54.755
If someone physically takes my laptop,

00:01:54.755 --> 00:01:59.163
the data on that laptop is safe because it's physically encrypted.

00:01:59.163 --> 00:02:02.375
If your laptop, I mean if I take your laptop right now while it's powered on,

00:02:02.375 --> 00:02:04.478
while it's logged in?

00:02:04.478 --> 00:02:05.486
I've had it.

00:02:05.486 --> 00:02:07.233
Right, yeah, that's not so useful.

00:02:07.233 --> 00:02:10.401
When I power my laptop off and I put it into my bag

00:02:10.401 --> 00:02:11.673
and I'm walking down the street,

00:02:11.673 --> 00:02:13.735
if someone koshes me over the head with a sock full

00:02:13.735 --> 00:02:18.276
of quarters and takes my bag, the disk encryption is engaged,

00:02:18.276 --> 00:02:22.162
and by the time you crack the crypto key to get at that,

00:02:22.162 --> 00:02:24.555
I'm going to find you, right?

00:02:24.555 --> 00:02:26.264
And I'm not going to be very happy about it.

00:02:26.264 --> 00:02:28.543
So disk encryption is great for protection of

00:02:28.543 --> 00:02:32.830
physical assets that may be stolen, but what about logical assets?

00:02:32.830 --> 00:02:39.419
Well, so if it may be stolen and the disk encryption is engaged at the time,

00:02:39.419 --> 00:02:39.952
right,

00:02:39.952 --> 00:02:41.743
this is the thing I have to harp at for my colleagues when we talk

00:02:41.743 --> 00:02:45.008
about our own information security practice,

00:02:45.008 --> 00:02:45.310
power this off.

00:02:45.310 --> 00:02:50.950
If you just put it in sleep and throw it in your trunk, it's not helpful.

00:02:50.950 --> 00:02:51.896
It's decrypted.

00:02:51.896 --> 00:02:55.489
A server or a disk storage array or something like that

00:02:55.489 --> 00:02:57.972
that's relying on disk or volume encryption,

00:02:57.972 --> 00:03:02.603
it has to decrypt that to use it, right?

00:03:02.603 --> 00:03:03.646
So the server that's always on,

00:03:03.646 --> 00:03:07.041
except for those infinitesimal moments where it's rebooting,

00:03:07.041 --> 00:03:10.031
right, is always in a decrypted state.

00:03:10.031 --> 00:03:13.200
So does that meet the requirement in the standard to

00:03:13.200 --> 00:03:15.411
encrypt cardholder data in storage?

00:03:15.411 --> 00:03:18.680
So many organizations would argue that it does,

00:03:18.680 --> 00:03:19.419
right,

00:03:19.419 --> 00:03:24.388
because there's no provision in 3.4 .1 where it talks about disk

00:03:24.388 --> 00:03:29.377
encryption that limits it to those particular states and so there are a

00:03:29.377 --> 00:03:31.692
lot of organizations that sort of lean on that and say,

00:03:31.692 --> 00:03:34.175
well, at least we're doing disk encryption,

00:03:34.175 --> 00:03:38.166
and my argument is, but it doesn't do anything,

00:03:38.166 --> 00:03:38.732
right?

00:03:38.732 --> 00:03:42.728
And there's another sort of variant on that,

00:03:42.728 --> 00:03:47.075
right, where there's like whole database encryption?

00:03:47.075 --> 00:03:47.855
Transparent data.

00:03:47.855 --> 00:03:49.479
Transparent data encryption or where you have the

00:03:49.479 --> 00:03:51.580
appliance sort of sitting in front of,

00:03:51.580 --> 00:03:54.583
you know doing the cryptographic action kind of back

00:03:54.583 --> 00:04:00.217
and forth in front of the database, and then you know,

00:04:00.217 --> 00:04:02.447
then that turns a cryptography problem into an access control problem.

00:04:02.447 --> 00:04:04.622
If your access control is insufficiently granular,

00:04:04.622 --> 00:04:09.871
right, if everybody can query that database and get the decrypted values,

00:04:09.871 --> 00:04:10.292
you know,

00:04:10.292 --> 00:04:12.751
now you've got a hardware device that's burning electricity and

00:04:12.751 --> 00:04:16.339
warming the planet and not really doing any good.

00:04:16.339 --> 00:04:16.771
So again,

00:04:16.771 --> 00:04:20.413
the message there is that the cryptography might not fix your security problem.

00:04:20.413 --> 00:04:22.329
You might be able to get it past an assessor,

00:04:22.329 --> 00:04:25.680
a poor assessor would allow disk cryptography,

00:04:25.680 --> 00:04:26.376
yeah?

00:04:26.376 --> 00:04:27.332
Yeah, so well,

00:04:27.332 --> 00:04:36.245
and what it tells you is that good cryptography doesn't fix bad architecture.
