Penetration Test Assistance

Mailman List Mirror (Read Only)
1

Howdy all,

I'm a Security Manager of a large network, we are conducting a Pentest next month and the testers are demanding a complete network diagram of the entire network. We don't have a "complete" network diagram that shows everything and everywhere we are. At most we have a bunch of network diagrams that show what we have in various areas throughout the country. I've been asking the network engineers for over a month and they seem to be too lazy to put it together or they have no idea where everything is.

I've never been in this situation before. Should I be honest to the testers and tell them here is what we have, we aren't sure if it's accurate; find everything else? How would they access those areas that we haven't identified? How can I give them access to stuff that I didn't know existed?

What do you all do with your large networks? One huge network diagram, a bunch of network diagrams separated by region, or both? Any pentest horror stories?

Thanks,

Tim

2

Any penetration test should only require your networks and masks. As
far as a diagram it is of value to keep a staff member with the
singular task of documentation and auditing or an optional contract
basis. Small things like typographical errors can cause great
confusion in emergency situations. Take the time and do it right. I
personally prefer the flexibility and ease of use that Mediawiki
offers but other free and pay solutions exist.

3

I'm a Security Manager of a large network, we are conducting a Pentest next month and the testers are demanding a complete network diagram of the entire network. We don't have a "complete" network diagram that shows everything and everywhere we are. At most we have a bunch of network diagrams that show what we have in various areas throughout the country. I've been asking the network engineers for over a month and they seem to be too lazy to put it together or they have no idea where everything is.

As someone who is charged with both engineering and maintaining the records and diagrams of a large network, I take exception to the word 'lazy' :wink: Network engineers tend to be an over-worked lot, and their work is often interrupt-driven, so large blocks of time to work on a single task are often a rarity.

The issue is that if they haven't kept their diagrams up to date (many people don't, unfortunately), then getting them up to date turns into a much more labor-intensive job. If they have kept the diagrams up to date and they're just not getting them to you, then take the issue up with their manager.

There might also be the question of how much information they are allowed to release to third parties, even if it is for a pentest. This could mean that some information might need to be removed or redacted from the diagrams. Again, the engineering manager/director/CIO/CTO might be able to provide clarification on this.

I've never been in this situation before. Should I be honest to the testers and tell them here is what we have, we aren't sure if it's accurate; find everything else? How would they access those areas that we haven't identified? How can I give them access to stuff that I didn't know existed?

From what I've seen, in-depth pentests are often done in coordination with

other groups, such as engineering/ops. In a large network, that's often done out of necessity, if for no other reason than dealing with issues like the ones you've raised (logistics, communication, etc...).

What do you all do with your large networks? One huge network diagram, a bunch of network diagrams separated by region, or both? Any pentest horror stories?

I don't have any pentest horror stories, but sometimes large network diagrams have to be broken up into pieces, to maintain some degree of readability. Large diagrams can get cluttered very quickly if you try to put every minute piece of detail on them. I tend to treat the main diagram as a high-level view of the network, and then either break out sections that need more detail as a separate drawing, or as a link to our internal knowledge base that can go into very high detail, including pictures, access information, etc.

There is no right way to diagram every network. It depends on what best suits your needs, and what established proceures are already in place.

jms

4

A complete diagram makes their life easier, may make for a more
complete test, but they are working for you, so if you don't have it,
you don't have. I'm not a big fan of having a single diagram with
everything laid out anyway, but I'm from the old shcool.

-jim

5

Howdy all,

I'm a Security Manager of a large network, we are conducting a
Pentest next month and the testers are demanding a complete network
diagram of the entire network. We don't have a "complete" network
diagram that shows everything and everywhere we are. At most we have
a bunch of network diagrams that show what we have in various areas
throughout the country. I've been asking the network engineers for
over a month and they seem to be too lazy to put it together or they
have no idea where everything is.

I've never been in this situation before. Should I be honest to the
testers and tell them here is what we have, we aren't sure if it's
accurate; find everything else? How would they access those areas
that we haven't identified? How can I give them access to stuff
that I didn't know existed?

What do you all do with your large networks? One huge network
diagram, a bunch of network diagrams separated by region, or both?
Any pentest horror stories?

Logical diagrams tend to elide the information consider unnecessary for
them to be suitably informative.

An ethernet switch with 560 network segments radiating out from it may
be accurate but not all that easy to parse or use.

Documentation needs to be sufficiently accurate and appropiate to the
tasks at hand, so it may be that you don't have what you need or perhaps
you do.

6

It's not much of a penetration test, imho, if the "attackers" have detailed
knowledge of your network and systems before the attack. You should
determine what kind of a scenario you are trying to simulate, and how the
results will be used to improve security. Is this a "black box" situation,
where you want to see what potential attackers can discover about your
systems without insider information? Or will this be a step by step,
examine each part of the system and then step back to see what's going on
from a high level scenario?

If you're trying to both reduce vulnerabilities and your attack profile, I
would go for the black box approach and see what your pentesters can come
up with themselves. Man is a resourceful creature, and you never know what
they could turn up.

Q

7

Not discounting the need for network diagrams, there are also differing approaches to pen testing. One alternative is a sort of black-box approach where the pen testers are given little or no advanced knowledge of the network. It is up to them to 'discover' what they can through open source means and commence their attacks from what they glean from their intelligence gathering. This way they are realistically mimicking the hacker methodology.

Ron Baklarz C|CISO, CISSP, CISA, CISM, NSA-IAM/IEM
Chief Information Security Officer
Export Control Compliance Officer
National Passenger Railroad Corporation (AMTRAK)
10 G Street, NE Office 6E606
Washington, DC 20002
BaklarR@Amtrak.com

8

Yup, a list of subnets in use on your network is all I've ever needed to provide to pen testers in the past on the few occasions I've worked with them. A good pen test should scan everything on your network anyways, with a reasonable chance of figuring out what everything is.

As far as horror stories... yeah. My most memorable experience was a guy (with a CISSP designation, working for a company who came highly recommended) who:
     - Spent a day trying to get his Backtrack CD to "work properly". When I looked at it, it was just a color depth issue in X that took about 45 seconds from "why is this broken?" to "hey look, I fixed it!".
     - Completely missed the honeypot machine I set up for the test. I had logs from the machine showing that his scanning had hit the machine and had found several of the vulnerabilities, but the entire machine was absent from the report.
     - Called us complaining that a certain behavior that "he'd never seen before" was happening when he tried to nmap our network. The "certain behavior" was a firewall with some IPS functionality, along with him not knowing how to read nmap output.
     - Completely messed up the report -- three times. His report had the wrong ports & vulnerabilities listed on the wrong IPs, so according to the report, we apparently had FreeBSD boxes running IOS or MS SQL...
     - Stopped taking our calls when we asked why the honeypot machine was completely missing from the report.

In general, my experience with most "pen testers" is a severe disappointment, and isn't anything that couldn't be done in-house by taking the person in your department who has the most ingrained hacker/geek personality, giving them Nessus/Metasploit/nmap/etc, pizza and a big ass pot of coffee, and saying "Find stuff we don't know about. Go.". There is the occasional pen tester who is absolutely phenomenal and does the job properly (i.e. the guys who actually write their own shellcode, etc), but the vast majority of "pen testers" just use automated tools and call it a day. Like everything else in IT, security has been "commercialized" to the point where finding really good vendors/people is hard, because everyone and their mom has CEH, CISSP, and whatever other alphabet soup certifications you can imagine.

9

Tim,

Your system is what it is, including any defects in configuration
management. Provide the testers with what you have, give them contact
info for the engineers so they can ask questions and specify that you
expect strengths and weaknesses in configuration management which
impact system security to be reflected in their report.

Regards,
Bill Herrin

10

I'd treat this as the first of their pen tests - a social engineering
attack to obtain secret information about the network, and refuse.

Aled

11

There are definitely a number of incredible pen-testers out there. But I agree with Peter… If you end up with a "report" that's nothing more than an executive statement pasted at the top of a Nessus report, then you've wasted your money. To be honest, I'd recommend getting a sample report from the company and quiz them on it before committing to a contract with them.

12

Hi Tim,

A _good_ pen test team would not need a network diagram. Their first round of penetration test would have them build their own network diagram from their analysis of your network.

Barry

13

Seriously.

--p

14

I'm with Barry--a network diagram showing everything from the pov of the pen team should be part of the end report.

--p

15

There are lots of reasons why a pentester would want a network diagram.

The foremost being a point to which they can say, these are the networks that I was given as a point of reference to pentest.

This is often a CYA policy for when people start complaining about the scanning that is going to occur and potentially break their systems.

Cheers,
Harry

16

The bit of information that's missing here is what are you trying
to pentest, and by extension how much do you want to pay your pentest
firm?

For some folks a pentest means starting with zero information and
trying to get IP packets passed a firewall or IDS's undetected.
Basically pentesting layer 3 infrastructure.

For other folks a pentest is purely an application level exercise,
you give the pentester an account on your customer portal for
instance, a full network diagram, and let them try things like SQL
injection or cross site scripting at the applications layer.

Your pentest firm can start with zero information and work all the
way up to an application level attack, but that's costly and time
consuming. Providing them some information is a way to short circuit
the process.

If you (or appropriate company representative) haven't already
discussed the pros and cons with your pentest firm you're off on the
wrong foot.

17

Tim,

In the past I've used high level diagrams to illustrate the overall network topology with individual tabs (drill down) per data center or POP.
The first step to assessing risk is to identify your assets. I'd suggest performing a discovery of your network. Keep in mind Pen tests are typically inconclusive of availability based threats DOS/DDOS (a very high risk today) and in fact specifically avoid tests which might cause degradation of service. I'd suggest including volumetric network (tcp, udp), application floods (http get, post, etc. /dns query floods, etc.) and slow and low attacks.

Best of Luck,

Dennis

18

I agree with a lot of what you've said, but there are absolutely good security guys (pen tester, vulnerability assessors, etc) that use both open source and commercial automated tools, but still do a fantastic job because they understand the underlying technologies and protocols.

I used to do a lot of this in the past, had lots of automated tools, and only occasionally wrote some assessment modules or exploit code if necessary.

But again, a person in that position has to understand technology holistically (network, systems, software, protocols, etc).

-b

19

You should have a look at the Pentest Standards page, it was created
by some very skilled Pen Testers how are trying to create a minimum
standard for all tests and reporting.

http://www.pentest-standard.org/index.php/Main_Page

Also you should just have to give them your external net-block
allocation that is in scope unless it is a more forced test and not a
general external test.

20

I completely agree. I didn't mean to imply that using automated tools is a bad thing -- simply that running an automated tool to pump out a report with no further investigation isn't really a useful pen test. I've seen vendors whose "comprehensive penetration testing" was basically "We'll run Nessus against your network, write up an executive summary and email you the scan results. Quite the bargain for $20K!"

Automated tools are definitely good to provide a first pass over a network, but even then multiple tools should be used, and an experienced eye should review the results for anomalies (whether that's a vulnerability that has a chance for false positives, discrepancies between the results of two or more automated tools, etc). That kind of work, along with more aggressive pen tests and exploit development, need a "guru meditation"-level understanding of the involved technologies, protocols, etc, as you mentioned.

Like everything else IT, the specific tools used are more or less immaterial to an excellent practitioner -- a good programmer can hack code in any language, a good network engineer can use any brand of network equipment, etc -- because these types of people truly understand the systems they're dealing with, and use tools to accomplish a specific task which fits into part of the "big picture" they have in their heads. Poor practitioners in a field use tools for the sake of using the tool ("I'm scanning a network with Nessus because that's what the certification course told me to do") without that deep level of understanding, and therefore don't provide any real value to the process.

- Pete