On 01/22/2014 06:55 AM, Paul Robert Marino wrote:
James
thats a great analysis I highly encourage you to write an article on
the subject some time.

one more thing I would add is about authentication. If you are planing
to use centeralized authentication make sure you do a full LDAP 3
implementation with TLS and Kerberos 5. do not use the standard config
most people use where they put a password in the LDAP client
configuration always use Kerberos Keytab files and secure them well.

I would also highly suggest a password auditing system esspecially if
you dont intend to use centralized authentication. There is a decent
one thats not to expensive Ive used in the past which also acts as a
password vault called Password Manager Pro from Manage Engine.
Unfortunately there are no free alternative out there which surprises
me because it wouldn't be too difficult a project to write. Again that
should be in its own isolated local network that is not accessible
from any where but a few internal workstations.




On Wed, Jan 22, 2014 at 6:11 AM, James Rogers <wa...@preternatural.net> wrote:
Just one more thing: you must run your datastore of card holder information
on server separate from your external interface. The cardholder datastore
should be accessible only via a local (non-routable) network. Ideally, you
should mac address restrict this using IPTables on the machine that holds
'things'. Never store CVV codes, although you might want to discuss with
your client the relative benefits of CVV codes vs address verification.
Gateways charge more for CVV/CVF. You might want to request the code, never
store it, and check it only according to internal constraints, but always
perform address verification.




On Wed, Jan 22, 2014 at 5:40 AM, James Rogers <wa...@preternatural.net>
wrote:

You'll need an application firewall. If you're using Apache, mod_sec will
work. Put up a proxy and filter connections. Don't run the proxy on the same
machine (VM or HM) as your app and/or its storage if you can manage.

Likely this means running a separate VM/HM in front of your web app and
that acts as a scanning proxy running mod_sec.

You should also run a HID on all machines and an NID on your border
firewalls. Pick people from your client's execs to send the warnings &
reports to (not the same person)  as you will need to list them in your PCI
docs, along with a _responsible_ tech who actually pays attention at 4AM.

As far as HID's: Tripwire is venerable, AIDE is current from my
understanding. You might also check into Beltaine/Lucifer.

And NID: SNORT or Suricata. And if you feel brave / if you need it: feed
the output of your NID into iptables for an active firewall. If anyone trips
you're HID, it's kind of baby vs bath-water time anyway once you have it
tuned: they're in... what do you do. Always leave some trips around that let
people know even if it is a rarely occurring legitimate changes. Testing the
alarms regularly is a part of the alarm system.

Unless you're _providing_ PCI compliance to your client as a documented
service, you should ask them for their requirements. In other words, don't
eat more liability than you need to. Unless you're a lawyer, then you will
have separate ethical requirements.

This is vague (certainly not legal) advice, give more on your requirements
and/or seek a lawyer.


On Wed, Jan 22, 2014 at 5:10 AM, James Rogers <wa...@preternatural.net>
wrote:

PCI compliance is largely related to what PCI level your client is at.
That level is related to how much money they move each year.

Selinux (or Apparmor) is good. Some sort of MDAC on your machines that
handle PIF is a good thing, but as you noted, it won't protect you from
social hacks, just from the chaff spewed on the internet by C2 servers and
their botnets.

If you don't find it too onerous, encrypt the swap and the filesystem. Be
aware of the dangers of this before you start and plan for them. Have
safe-houses your client plans and pays for that store the relevant
information. Use M-Disks to store it? And encrypted drives.  You'll know
what to do once you explore the dangers of encrypted filesystems, and your
client will produce locations.

As far as AV... hmmm... I would go with 3 engines of your choice, one of
which should be ClamAV. I would go with Frisk/F-Prot as the next (they're
not expensive). And then maybe sophos if your clients have the cash to
spend. What you're largely looking at from the AV scanners is that they
protect the people visiting your site. Unless you're doing something with
the DoD and then you will have different requirements.

The next place to look (or the first even) will be an active daily
scanner for your external reporting. If you're dealing with a Merchant Bank
/ Acquiring Bank, use theirs as that will be least expensive. Otherwise...
Hackersafe/MCafee is a reasonable choice as it is automated and you don't
have to deal with people very often; they're owned by Intel so they're not
going to dry up and blow away, which is a plus. You should be doing their
job beforehand using nessus/something else. Your external scanner will give
you a badge to display. Basically, the scanning company will run a port and
vulnerability scan and then offer you remediation recommendations and
requirements. If you don't solve your problems, you lose their seal on your
site.

Every year you will need to forward PDF reports from the company you
contract to scan you to your merchant bank and any other parties that
require PCI compliance. It's not a big thing, but something that must be
done, and you will need to find the contact information for the people
involved and make your client aware that they need to pay attention to it
and keep track of any change in contacts after your contract expires.
Remember to charge for the time you spend on this. Contractors often forget
to charge for doing small things, and so they don't get done. Make a point
to charge your client and provide the information they need to keep doing
business.

This is likely more than you wanted.





On Tue, Jan 21, 2014 at 12:39 AM, ToddAndMargo <toddandma...@zoho.com>
wrote:

Hi All,

    I am in the thinking phase of a new server for
a customer.  The server needs to be PCI Compliant
(credit card security).  PCI is really a huge
paper chase and although it adds a lot of good
practices, it doesn't really address the human
factor like it should, which is where most of the
breaches come these days.

    I was going to suffer with SE Linux left on.
Samba with SE Linux: I will say a few blue
words before it is over.  :'(

    I have the File Integrity Software picked out
(CimTrak) as I has used it in the Windows Arena
and like how it works.  And the sales and tech support
is astounding.

    What I have yet to pick out is an Anti Virus (AV).
It is part of the paper chase.  Looking over at

    http://chart.av-comparatives.org/chart1.php

I am not seeing Clam AV.  I know Kaspersky has one,
but the last time I tried it, it was a mess.
Any thoughts on an AV?

   If you look at the chart, no one did worse than
M$ Security Essentials in December.  Chuckle.

Many thanks,
-T

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Computers are like air conditioners.
They malfunction when you open windows
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~






Thank you!

--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Computers are like air conditioners.
They malfunction when you open windows
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Reply via email to