WEBVTT

00:00:00.105 --> 00:00:02.176
So the rest of requirement 3.

00:00:02.176 --> 00:00:04.761
We've done the hard bit about let's not keep cardholder data,

00:00:04.761 --> 00:00:06.914
let's not keep sensitive authentication data.

00:00:06.914 --> 00:00:09.515
Now we're in Kansas,

00:00:09.515 --> 00:00:15.685
and we are actually deciding as a business we have a legitimate business need to

00:00:15.685 --> 00:00:18.982
store primary account numbers as PANs and cardholder data.

00:00:18.982 --> 00:00:19.839
Right.

00:00:19.839 --> 00:00:20.875
So now we need to protect them.

00:00:20.875 --> 00:00:21.794
Yes.

00:00:21.794 --> 00:00:23.443
So I want to think about this really in two ways.

00:00:23.443 --> 00:00:26.847
The first bit is how are we going to protect them?

00:00:26.847 --> 00:00:27.638
Right.

00:00:27.638 --> 00:00:31.868
And then the second bit is okay, now we're going to use cryptography,

00:00:31.868 --> 00:00:34.535
and then we'll talk about the cryptography stuff.

00:00:34.535 --> 00:00:36.231
Sure.

00:00:36.231 --> 00:00:38.187
So DSS gives us a few options, doesn't it?

00:00:38.187 --> 00:00:38.870
Yes.

00:00:38.870 --> 00:00:40.042
It talks about we could truncate them?

00:00:40.042 --> 00:00:41.085
Yes.

00:00:41.085 --> 00:00:43.470
It talks about the fact we could encrypt them,

00:00:43.470 --> 00:00:45.721
and then it talks about tokenization as well.

00:00:45.721 --> 00:00:46.868
Yes.

00:00:46.868 --> 00:00:48.860
So let's run through very quickly the merits of those.

00:00:48.860 --> 00:00:49.865
Sure.

00:00:49.865 --> 00:00:53.717
So from a risk perspective, as well as your job as an assessor,

00:00:53.717 --> 00:00:57.280
you know, obviously I'm guessing that truncation is the easiest to assess?

00:00:57.280 --> 00:00:58.960
Truncation is the easiest to assess, right,

00:00:58.960 --> 00:01:03.182
because at that point you're getting rid of what's technically

00:01:03.182 --> 00:01:06.905
cardholder data and sort of rendering it down into,

00:01:06.905 --> 00:01:09.817
you know, at most first six, last four.

00:01:09.817 --> 00:01:12.782
So you're doing the Jedi mind trick?

00:01:12.782 --> 00:01:15.227
This is not the cardholder data you're looking for?

00:01:15.227 --> 00:01:15.726
Yeah, well,

00:01:15.726 --> 00:01:18.663
I mean you're doing it and then actually like excising some

00:01:18.663 --> 00:01:22.918
of the digits out and throwing them away.

00:01:22.918 --> 00:01:25.044
So actually it's not a Jedi mind trick, it's actually magic.

00:01:25.044 --> 00:01:25.691
Right.

00:01:25.691 --> 00:01:30.460
Right, we've taken cardholder data, we've taken the middle digits out of it,

00:01:30.460 --> 00:01:32.026
and we've turned it into not cardholder data.

00:01:32.026 --> 00:01:32.839
Right, right, right,

00:01:32.839 --> 00:01:36.693
and so now what you've got is just this limited set of

00:01:36.693 --> 00:01:39.107
digits that's not the full number.

00:01:39.107 --> 00:01:39.423
Okay.

00:01:39.423 --> 00:01:43.365
How many, remind me, how many digits do I have to take out of the middle?

00:01:43.365 --> 00:01:48.643
Right, so you can keep what they describe as the first six and the last four,

00:01:48.643 --> 00:01:48.991
right?

00:01:48.991 --> 00:01:55.545
So usually that's eight digits in the middle, am I counting right?

00:01:55.545 --> 00:01:55.907
Six.

00:01:55.907 --> 00:01:57.211
Six digits in the middle,

00:01:57.211 --> 00:02:01.557
maybe five for American Express cards that you're pulling

00:02:01.557 --> 00:02:03.350
out of the middle and getting rid of, right?

00:02:03.350 --> 00:02:07.406
And you could do less, you could just keep the last four.

00:02:07.406 --> 00:02:08.066
You could maybe,

00:02:08.066 --> 00:02:11.568
I've seen like the first two and the last four or

00:02:11.568 --> 00:02:13.457
similar configurations like that.

00:02:13.457 --> 00:02:18.710
So if you've got some need for the last four for verification purposes,

00:02:18.710 --> 00:02:19.168
right,

00:02:19.168 --> 00:02:21.233
or you need the first two to identify the card type or

00:02:21.233 --> 00:02:27.084
you need the first six for the bin, you know then you can do that,

00:02:27.084 --> 00:02:29.234
but then the stuff in the middle, you get rid of,

00:02:29.234 --> 00:02:30.145
and that's that.

00:02:30.145 --> 00:02:32.836
And that's great because now as an assessor it's just truncating.

00:02:32.836 --> 00:02:35.949
You can check in a database, you're seeing truncated card numbers.

00:02:35.949 --> 00:02:38.699
Right, all these truncated card numbers and that's it.

00:02:38.699 --> 00:02:38.965
Okay,

00:02:38.965 --> 00:02:41.997
and actually we should talk a bit more about truncation because

00:02:41.997 --> 00:02:44.872
there is a really good FAQ because now people are making card

00:02:44.872 --> 00:02:46.729
numbers that are more than sixteen digits.

00:02:46.729 --> 00:02:47.656
Sure.

00:02:47.656 --> 00:02:48.684
In Europe we've had Maestro cards,

00:02:48.684 --> 00:02:51.029
which have been longer than 16 digits for a while,

00:02:51.029 --> 00:02:53.922
and now I know where're running out of card numbers,

00:02:53.922 --> 00:02:57.319
the card brands are talking about adding and having longer numbers.

00:02:57.319 --> 00:02:58.004
Right.

00:02:58.004 --> 00:03:01.272
So there is an FAQ on the PCI SSC website,

00:03:01.272 --> 00:03:06.558
which actually goes into great detail about how we can

00:03:06.558 --> 00:03:08.866
truncate cards and what different brands allow,

00:03:08.866 --> 00:03:11.057
because these are actually defined by the brand.

00:03:11.057 --> 00:03:15.025
This is one of the weird things where the PCI,

00:03:15.025 --> 00:03:17.036
where the brands write this bit for the PCI Council.

00:03:17.036 --> 00:03:18.381
Right.

00:03:18.381 --> 00:03:19.964
Because they've all got different rules about what numbers you

00:03:19.964 --> 00:03:21.753
can keep and what numbers you can't keep.

00:03:21.753 --> 00:03:24.472
So this FAQ is like the one place to look.

00:03:24.472 --> 00:03:26.019
I guess you look at this as an assessor.

00:03:26.019 --> 00:03:27.115
Right.

00:03:27.115 --> 00:03:29.748
If you see someone with a 19-digit PAN, to refresh yourself about,

00:03:29.748 --> 00:03:31.473
what do I do here?

00:03:31.473 --> 00:03:34.131
Yeah, I know I used to like print this one out and carry it with me.

00:03:34.131 --> 00:03:36.861
Yeah, that's probably a good idea.

00:03:36.861 --> 00:03:43.618
Okay, that's truncation, now the next thing is tokenization.

00:03:43.618 --> 00:03:46.910
Right.

00:03:46.910 --> 00:03:50.669
I mean, talk to me about really, as an assessor,

00:03:50.669 --> 00:03:53.338
what are you looking for to makes sure a token is a token?

00:03:53.338 --> 00:03:56.017
Because that does strike me as a potential Jedi mind trick.

00:03:56.017 --> 00:03:59.346
That has the potential to be more smoke and mirrors

00:03:59.346 --> 00:04:02.689
than sort of useful substance,

00:04:02.689 --> 00:04:05.050
and tokenization is one of those things that means slightly

00:04:05.050 --> 00:04:10.296
different things in certain contexts.

00:04:10.296 --> 00:04:12.986
So we're not talking about tokenization on devices like Apple

00:04:12.986 --> 00:04:15.059
Pay or Android Pay or anything like that?

00:04:15.059 --> 00:04:17.164
Correct, right, let's put that aside, that's something else.

00:04:17.164 --> 00:04:19.691
This is a merchant or a service provider replacing the

00:04:19.691 --> 00:04:23.046
card with something that's not the card?

00:04:23.046 --> 00:04:26.177
Right, so then the question, the first question is,

00:04:26.177 --> 00:04:29.050
where do you get the tokens, right?

00:04:29.050 --> 00:04:34.388
If you or if the assessed entity is sort of doing its own tokenization,

00:04:34.388 --> 00:04:38.006
it's got some vault somewhere where there's a database with two

00:04:38.006 --> 00:04:40.862
columns and a PAN on one side and a token on the other,

00:04:40.862 --> 00:04:41.505
I mean,

00:04:41.505 --> 00:04:46.720
then it still has cardholder data stored and it still needs to

00:04:46.720 --> 00:04:49.353
protect that according to what's in requirement 3,

00:04:49.353 --> 00:04:56.032
and what that maybe does is then the database or some other system that's

00:04:56.032 --> 00:05:01.263
using the token maybe even doesn't have those requirements about stored

00:05:01.263 --> 00:05:04.276
cardholder data applied to those systems.

00:05:04.276 --> 00:05:04.864
Yeah.

00:05:04.864 --> 00:05:05.161
Right?

00:05:05.161 --> 00:05:06.647
And maybe that's a benefit?

00:05:06.647 --> 00:05:07.915
Yeah.

00:05:07.915 --> 00:05:08.150
Good!

00:05:08.150 --> 00:05:08.384
Right?

00:05:08.384 --> 00:05:09.089
Well and good,

00:05:09.089 --> 00:05:12.742
but you still have that sort of token vault that you have to contend with,

00:05:12.742 --> 00:05:13.278
right?

00:05:13.278 --> 00:05:14.396
I think that's more of a minority case.

00:05:14.396 --> 00:05:18.189
The more likely cases, the payment processing entity,

00:05:18.189 --> 00:05:26.156
sort of upstream from that organization gets the PAN from,

00:05:26.156 --> 00:05:29.193
you know, the merchant or from the service provider or from,

00:05:29.193 --> 00:05:31.436
you know, whoever the entity is.

00:05:31.436 --> 00:05:33.978
Okay, so that's your payment service provider or your merchant acquiring bank?

00:05:33.978 --> 00:05:36.552
Right, so you send the PAN as part of the transaction?

00:05:36.552 --> 00:05:37.729
Right.

00:05:37.729 --> 00:05:38.774
And they give you a token?

00:05:38.774 --> 00:05:39.857
And they give you a token back.

00:05:39.857 --> 00:05:43.101
Right, and then you can use that token to do what?

00:05:43.101 --> 00:05:44.844
That depends.

00:05:44.844 --> 00:05:45.767
Right.

00:05:45.767 --> 00:05:47.073
So in some cases,

00:05:47.073 --> 00:05:51.136
that token could potentially be used for recurring transactions?

00:05:51.136 --> 00:05:52.956
Right.

00:05:52.956 --> 00:05:57.426
And then I think you have some questions about is that

00:05:57.426 --> 00:05:59.705
a quote-unquote high-value token?

00:05:59.705 --> 00:06:00.657
Right.

00:06:00.657 --> 00:06:02.382
If I'm the attacker and I capture that token,

00:06:02.382 --> 00:06:06.410
can I replay that and affect some fraud?

00:06:06.410 --> 00:06:07.165
So now we're talking risks rather than DSS because I'm

00:06:07.165 --> 00:06:08.950
going to ask you as an assessor,

00:06:08.950 --> 00:06:13.552
if you get a token back from your merchant acquiring bank or payment processor,

00:06:13.552 --> 00:06:15.347
is that out scope now?

00:06:15.347 --> 00:06:17.776
Is that a PAN or is that not a PAN?

00:06:17.776 --> 00:06:21.947
I would say it depends, because that's the assessor answer, right?

00:06:21.947 --> 00:06:27.101
But, so just pro tip, when the assessor says it depends,

00:06:27.101 --> 00:06:31.054
right, make the assessor tell you on what it depends,

00:06:31.054 --> 00:06:34.111
and that tells you if you've got a good assess or not.

00:06:34.111 --> 00:06:34.726
Okay.

00:06:34.726 --> 00:06:35.610
Right?

00:06:35.610 --> 00:06:45.138
So it depends; if the token is limited to just that particular transaction,

00:06:45.138 --> 00:06:50.024
right, such that it's not about that card number,

00:06:50.024 --> 00:06:54.240
but it's literally just about that transaction and all I can maybe then do is

00:06:54.240 --> 00:06:58.362
replay it to void that transaction or something like that,

00:06:58.362 --> 00:06:59.665
then I think your risk is pretty small.

00:06:59.665 --> 00:07:00.664
Yeah.

00:07:00.664 --> 00:07:03.393
If that, and you know, or for particular merchant,

00:07:03.393 --> 00:07:06.323
for that particular transaction, for that particular,

00:07:06.323 --> 00:07:08.864
you know, use, right, that particular dollar value,

00:07:08.864 --> 00:07:09.111
right?

00:07:09.111 --> 00:07:11.440
Then your risk is probably pretty small.

00:07:11.440 --> 00:07:14.317
If that token is good for that card number,

00:07:14.317 --> 00:07:18.939
for that PAN and is not necessarily restricted to a

00:07:18.939 --> 00:07:21.170
particular merchant or a particular transaction,

00:07:21.170 --> 00:07:21.349
right,

00:07:21.349 --> 00:07:26.552
and it could be replayed to go buy flat screen TVs from a big-box retailer,

00:07:26.552 --> 00:07:27.016
right?

00:07:27.016 --> 00:07:29.478
Now you've got a different risk characteristic.

00:07:29.478 --> 00:07:30.368
So it's meant to be a PAN?

00:07:30.368 --> 00:07:32.109
Right, and it really is.

00:07:32.109 --> 00:07:34.460
My take was always also this, was like,

00:07:34.460 --> 00:07:38.482
if the merchant can send the token to wherever it

00:07:38.482 --> 00:07:42.226
got it from and the PAN back Yes!

00:07:42.226 --> 00:07:42.597
Right?

00:07:42.597 --> 00:07:44.611
That's very reversible and very easy,

00:07:44.611 --> 00:07:46.303
how do they protect that and how do they protect?

00:07:46.303 --> 00:07:50.676
So we're actually moving away from DSS and into helping people be secure, yeah?

00:07:50.676 --> 00:07:51.600
Right, right, well,

00:07:51.600 --> 00:07:55.678
and I'd say then if you've got a situation where you can

00:07:55.678 --> 00:07:58.237
send the token back and get the PAN back,

00:07:58.237 --> 00:08:02.194
right, then we at least have to test the,

00:08:02.194 --> 00:08:03.896
okay, so now you get the PAN back, what happens,

00:08:03.896 --> 00:08:04.412
right?

00:08:04.412 --> 00:08:05.544
Does it get stored?

00:08:05.544 --> 00:08:07.874
What system does it transit through?

00:08:07.874 --> 00:08:08.657
Yeah.

00:08:08.657 --> 00:08:09.955
Now we think about that, right?

00:08:09.955 --> 00:08:12.978
And maybe those requirements apply in some capacity there.

00:08:12.978 --> 00:08:14.107
And really also where can that happen?

00:08:14.107 --> 00:08:17.279
So like if you have a payment switch that can do that,

00:08:17.279 --> 00:08:19.986
it's probably okay because that might be in a nice secure area,

00:08:19.986 --> 00:08:23.780
but if anyone in the organization can take that token and fire it

00:08:23.780 --> 00:08:26.092
off to an API somewhere and get the PAN back,

00:08:26.092 --> 00:08:27.052
that's probably not okay.

00:08:27.052 --> 00:08:28.958
That's less great, right?

00:08:28.958 --> 00:08:29.434
Yeah.

00:08:29.434 --> 00:08:31.014
Okay, yeah, you know,

00:08:31.014 --> 00:08:34.723
where I've recently been working we did a massive tokenization program,

00:08:34.723 --> 00:08:37.791
and it was culturally the right way of doing compliance

00:08:37.791 --> 00:08:40.583
for the organization of you know, you see a pattern.

00:08:40.583 --> 00:08:42.593
In my view, tokenization is really weird,

00:08:42.593 --> 00:08:45.354
is that as soon as, you need to tokenize as soon as you can,

00:08:45.354 --> 00:08:47.162
because there's no point tokenizing once you've brought

00:08:47.162 --> 00:08:48.661
half your environment into scope.

00:08:48.661 --> 00:08:49.698
Exactly.

00:08:49.698 --> 00:08:52.773
So your aim of tokenization is to tokenize as soon as you can see the card

00:08:52.773 --> 00:08:56.769
number and de-tokenize the latest possible moment you do it.

00:08:56.769 --> 00:08:58.799
Right.

00:08:58.799 --> 00:08:59.732
So you know,

00:08:59.732 --> 00:09:03.651
a lot of organizations will use devices like F5s and do

00:09:03.651 --> 00:09:05.366
tokenization on the way into the network.

00:09:05.366 --> 00:09:09.724
They decrypt the traffic, see the PAN, tokenize it straightaway.

00:09:09.724 --> 00:09:12.123
You can tokenize it even earlier from that using

00:09:12.123 --> 00:09:16.177
like content distribution networks, and they can tokenize it to CDN,

00:09:16.177 --> 00:09:19.421
and then tokenize on the way out, literally when you go for authorization.

00:09:19.421 --> 00:09:21.183
Because you know, if you're investing in tokenization,

00:09:21.183 --> 00:09:23.457
and I don't know if you agree with this,

00:09:23.457 --> 00:09:26.919
but my view is you should tokenize this as much as you can of your environment.

00:09:26.919 --> 00:09:28.777
Absolutely, yeah, if you're going to go down that road,

00:09:28.777 --> 00:09:31.431
if you're going to pay somebody for it, if you're going to,

00:09:31.431 --> 00:09:33.927
you know, get those sort of premium payment processing services,

00:09:33.927 --> 00:09:38.592
then you should derive the greatest benefit from it as possible.

00:09:38.592 --> 00:09:39.716
Tokens are great for two reasons.

00:09:39.716 --> 00:09:45.088
One is if a bad guy steals them, they can't do anything with them.

00:09:45.088 --> 00:09:48.182
So they're not valuable so they're not going to attack you for them,

00:09:48.182 --> 00:09:49.647
and secondly, it reduces your scope.

00:09:49.647 --> 00:09:51.342
And so we're always talking about compliance here,

00:09:51.342 --> 00:09:54.146
but tokens are great because they reduce your attack surface

00:09:54.146 --> 00:09:56.658
massively and bad guys won't come and take them from you.

00:09:56.658 --> 00:09:59.437
You know, it's not just about compliance.

00:09:59.437 --> 00:09:59.667
Right.

00:09:59.667 --> 00:10:01.923
So anything else we want to talk about tokenization before we move

00:10:01.923 --> 00:10:04.149
onto the thing that we don't want to talk about?

00:10:04.149 --> 00:10:04.399
Right,

00:10:04.399 --> 00:10:08.548
I'd say the only other thing to consider is just making sure

00:10:08.548 --> 00:10:13.596
that the tokens are not mathematically related to the PAN in

00:10:13.596 --> 00:10:15.769
some way that could be undone.

00:10:15.769 --> 00:10:18.898
So you can't like add 1 to a PAN and make it a token?

00:10:18.898 --> 00:10:20.844
That's probably not a good idea.

00:10:20.844 --> 00:10:22.533
Is there a standard for that?

00:10:22.533 --> 00:10:27.269
Is there a way you can evaluate tokenization appliances or tokenization systems?

00:10:27.269 --> 00:10:31.027
I believe there is, yes, I believe there is.

00:10:31.027 --> 00:10:33.895
I asked that question because I spent a lot time when I was on the Council,

00:10:33.895 --> 00:10:36.545
actually developing what turned out not to be a standard,

00:10:36.545 --> 00:10:39.197
but there's a guidance document, isn't there?

00:10:39.197 --> 00:10:44.384
Right, it's quite hard to find, but if you're doing your own tokenization,

00:10:44.384 --> 00:10:48.175
there's a really good document that describes how to do it well.

00:10:48.175 --> 00:10:48.973
Right.

00:10:48.973 --> 00:10:49.715
And do you use that when you're assessing?

00:10:49.715 --> 00:10:53.171
Do you say if a merchant says I've rolled my own tokenization system,

00:10:53.171 --> 00:10:56.380
is that one of the ways that you assess it?

00:10:56.380 --> 00:10:57.605
That would be what I would reach for,

00:10:57.605 --> 00:11:01.172
but usually what I find when there is some sort of

00:11:01.172 --> 00:11:04.786
tokenization effort where I can see both sides of it,

00:11:04.786 --> 00:11:09.727
right, is that there's essentially a stack of kind of numbers,

00:11:09.727 --> 00:11:12.346
sequential or randomly generated or whatever,

00:11:12.346 --> 00:11:15.317
but sort of not meaningfully related to the PAN,

00:11:15.317 --> 00:11:19.542
and then as PANs come in, they just take the next one,

00:11:19.542 --> 00:11:19.946
right?

00:11:19.946 --> 00:11:25.027
And so you know, are these numbers incremental to one another?

00:11:25.027 --> 00:11:30.027
Yeah, but that's just the history of the last PAN that came in, right?

00:11:30.027 --> 00:11:32.330
So that's very often what I see usually,

00:11:32.330 --> 00:11:35.625
because people don't want to figure out more complex algorithms.

00:11:35.625 --> 00:11:38.951
Yeah, my probably with some tokenization is some of it's done really badly.

00:11:38.951 --> 00:11:39.797
Yes.

00:11:39.797 --> 00:11:42.447
And you have to be really careful before you,

00:11:42.447 --> 00:11:42.826
I mean,

00:11:42.826 --> 00:11:45.588
this is one actually really good piece of advice; I hope you'll agree with this,

00:11:45.588 --> 00:11:47.746
is that if you're about to buy a tokenization system,

00:11:47.746 --> 00:11:50.754
get your QSA to be part of your RSA.

00:11:50.754 --> 00:11:51.588
Absolutely.

00:11:51.588 --> 00:11:52.423
Absolutely.

00:11:52.423 --> 00:11:54.418
Because the last thing you want is to buy a tokenization system,

00:11:54.418 --> 00:11:58.096
the assessor to come along and go, but they're not tokens.

00:11:58.096 --> 00:11:58.849
Right, they're not tokens, right?

00:11:58.849 --> 00:12:02.993
You've just xor'd this with a 1 or something stupid like that, right?

00:12:02.993 --> 00:12:11.850
You know, congratulations on dropping, you know, five figures on garbage, right?
