A Penetration Testing

Learning Kit
-Ariel Waissbein-
TROOPERS 08
Munich, Germany
April 24, 2008
 Objectives

• Present a methodology for teaching

pentesting.
– We take a holistic approach.
– Discuss what’s needed to teach pentesting.

• Introduce a pentesting simulation tool.

– Show how it covers many teaching necessities.

• Invite you to teach with this tool and discuss

future plans.
 Outline

1. A first lesson.
i. Pentesting history & motivation.
ii. A first attack, dissected.
2. Introducing the Learning Kit.
i. Functionalities of the simulation-attack suite.
ii. A glimpse on the simulation-attack suite.
3. Completing the lesson into a course.
4. Other applications for the kit.
5. Discussion.
 Our approach is to abstract highly technical
functions, e.g., treat them as APIs.
• Is there a way to learn pentesting gradually, without first being an
expert in:
– OS security.
– Network security.
– Cryptography.
– Access controls & authentication.
– Et cetera (Application security, wireless, voip, embedded devices,…).

• We’ll get there with our teaching methodology.

– designing attacks aided by a theoretical model,
– executing them with a pen-testing framework that abstracts
functionalities, and
– replacing real networks with simulations.

Disclaimer: Next we will make a pentest mockup. But I won’t demo… I’ll only show some figures. All of
you interested in testing the kit, shoot me an email and I’ll send you credentials to download a preliminary
version.
A FIRST PENTESTING CLASS
 Becoming a pen tester

This course introduces you into pentesting by letting you practice pentests.

Learn about the Pentest goals, Learn about

attacker’s motivations scoping and defense and how to
and objectives planing thwart it

Design and use Know the other Conduct complete

exploits and payloads attacker’s tools pentests

CISSPs know all this: Access control; Telecommunications and network security; Information
security and risk management; Application security; Cryptography; Security architecture and
design; Operations security; Business continuity and disaster recovery planning; Legal,
regulations, compliance and investigations; Physical (environmental) security.
LEARN SOME HISTORY FIRST
 We’ve grown from the MULTICS years

• Saltzer and Schroeder audited the MULTICS code for

security bugs in 1975.

• Within the military:

– Computer Security: the Achilles' heel of the electronic air force,
Schell ,1979.
– The orange book, 1985.

• Worms – the RTM worm hits in 1988.

• Hacking = adding a line in /etc/passwd.

• Enter firewalls (1989? or around that time).

• Go hack yourself!
– Improving the security of your site by breaking into it (1993) - Dan
Farmer, Wietse Venema.
I. Arce’s USENIX SECURITY `06 keynote talk/slides.
 Attackers are getting more pro
• Mitnick: IP spoofing & TCP hijacking (1994).

• DoS with TCP SYN flood attacks (1996).

• Buffer overflows get widely understood

– “Smashing the Stack for Fun and Profit,” AlephOne, Phrack 49,
1996.
– “Syscall Proxying,” M. Caceres. BlackHat 2002.
– “The shellcode generation,” I. Arce. IEEE S&P, Sept/Oct `04.

• Some nasty incidents:

– “An analysis of the Slapper Worm,” I. Arce & E. Levy, IEEE S&P
`03.
– Web-application worms: Santy, 2004.

See: J. Pincus, B. Baker , “Beyond Stack Smashing: Recent Advances

in Exploiting Buffer Overruns,” IEEE S&P, Jul/Aug `04
 A definition

A pentest is
a time-constrained effort
to evaluate the security of a target
using the enemy’s approach
 Much more than listing vulnerabilities

• Scoping
– Define an objective.
– Fix an adversarial model.
• Executing:
– Plan, execute the plan doing only what’s necessary & gain
access to important assets.
– Actions mimic attackers’ behavior, but must be robust.
– Document all that’s done.
• Reporting
– Presenting detailed technical info & risk analysis.
 Scoping

• The contract & the client

– Is in the finance space,
– Wants to have a perimeter audit,
– Will pay for two weeks time & 1 pen tester,
– Is specially concerned about data loss in an internal server and
some employee’s workstations.

Scope/Goal Clean up and

Definition reporting

Analysis &
Penetration
planning

Information Privilege
Attack
Gathering escalation

Gera Richarte “Modern Intrusion Practices,” BlackHat `03

 Planing
• The plan is to tackle internet-facing servers first.
– Next, attempt to pivot inside.
– Use new exploits!

• Let’s see how to do this in the next slide…

Scope/Goal Clean up and

Definition reporting

Analysis &
Penetration
planning

Information Privilege
Attack
Gathering escalation

Futoransky, Notarfrancesco, Richarte, Sarraute, “Building

Computer Network Attacks,” 2003
 Plan carefully to avoid spurious actions

• Most of what happens in a pentest can be described by actions that are

used for obtaining assets.
– We will only execute an action if we can profit from the result.
– The result of an action depends on the scenario characteristics, the already acquired
assets (and probably a probabilistic factor).

• A plan is a path in a graph describing a sequence of actions that must be

consecutively executed to get the asset “remote agent” in one of the target
machines.
Use WU-FTP
Use a banner grabber
globbing exploit
to detect OS
Detect OS
RedHat 7.1
Connected to Remote agent
Local Agent
200.177.22.12
Service
Probe IPs FTPd:21
200.177.22.0/8 Scan ports TCP22,
one at a time TCP465, TCP993, …
Futoransky, Notarfrancesco, Richarte, Sarraute “Building Computer
Network Attacks,” Corelabs Technical Report. 2003.
 Something like this should work!

Notes:
Although this attack is not likely to succeed today, it makes for a nice example.
 IG
• After some Intelligence gathering (passive) we find
web servers, name servers and mail servers.

• We use information gethering (passive & active) tools

to detect OSs and open UDP & TCP ports,
– looking for one computer at a time.
– only for known vulnerable services that we can exploit.

• The 2nd one is a RedHat 7.1 with TCP:21 open.

Scope/Goal Clean up and

Definition reporting

Analysis &
Penetration
planning

Information Privilege
Attack
Gathering escalation
 IG in details
• The user chooses one of a few tools for each step. All data
that’s found is combined and viewed by the user.

• Some Information gathering tools included are:

– Network discovery: ARP, TCP SYN packets, ICMP echo request, TCP
connect and passive discovery.
– Port scan.
– OS fingerprinting: Nmap OS stack fingerprinting, a home-brew OS
fingerpringting by neural networks tool.
– Or simply use one of GFI LANguard, Qualys, nmap, Nessus, PatchLink
STAT, and eEye Retina and import the results!
 Attack & privilege escalation
• We can detect vulnerabilities now!
– Say, because we know that RedHat 7.1 had a
vulnerable wu-ftp by default and port 21 is open.

• For each potential vulnerability discovered in IG, we

send an exploit.
– We send the exploit and wait for the exploit to work.
– In some cases, we might need to do some local IG and
execute a second exploit to gain root privileges.

Scope/Goal Clean up and

Definition reporting

Analysis &
Penetration
planning

Information Privilege
Attack
Gathering escalation
 An exploit is code that adds features
• Sometimes applications are developed or designed
with so-called security bugs.

• This means that attackers/users are provided with a

feature that wasn’t there by design. For example,
– A buffer overflow will typically give the attacker the ability to
execute arbitrary code in the compromised system.
– A SQL-injection vulnerability will give the attacker the ability
to execute queries in the underlying web application.
• Sometimes this will allow the pentester to compromise
the server in a second step.

Gera Richarte “Assembly for Exploit Writing,” course.

 Exploits are around the corner!
• As of today, there are several means to get exploits:
– Develop them on your own,
– Get a pen-testing product (Metasploit, Impact, Canvas, …),
– Buy them/download them on the internet.

Gera Richarte “Assembly for Exploit Writing,” course.

 Penetration
• The exploit is followed by a payload or “egg” that
installs an agent in the compromised system.

• We use the “(syscall proxying) agent” that will

provide:
– a stealth connection between agent and pen tester (I.e.,
much like a rootkit),
– “shellcode functionality,” and
– the ability to chain new agents.
Scope/Goal Clean up and
Definition reporting

Analysis &
Penetration
planning

Information Privilege
Attack
Gathering escalation
 Advances in payload engineering
• Pen testers must leverage exploits code to profit from
vulnerabilities:
– To this end they use the “new features.”
– For example, buffer overflows allow the pentester to inject code into a
new/running process. The code, is called an egg.
– Likewise a SQL-injection attack will allow the pentester to execute
~arbitrary SQL commands.

• Mastering these “egg injections” is what we call payload

engineering
– rx (read and exec).
– Loader payloads: MOSDEF (Canvas), InlineEgg (Gera).
– Meterpreter, VNC server, dll injection tool (Metasploit).
– polymorphic & factorized payload (K2).
– Syscall proxying (Core).

See: A. Waissbein, “Strong Payload Obfuscation and

Encryption,“ PacSec 2006
 Syscall proxying
• System calls or syscall are instructions issued by applications to the OS.
– Typically through libraries, such as libc.

• A syscall proxy mechanism enables a syscall client (running in one

computer) to send syscalls to a syscall server (running in a 2nd computer)
that executes them and returns answers.

• It works as if the penetration tester were executing code in the remote

computer, but he really is in the local computer.

M. Caceres, Syscall Proxying. BlackHat `02.

 Next, we pivot and restart the cycle
• The pen-tester has a new base camp!

• Continuing with the plan, he searches for internal

servers.

Scope/Goal Clean up and

Definition reporting

Analysis &
Penetration
planning

Information Privilege
Attack
Gathering escalation
 The pentester’s view is enlared
• Continuing with the plan, he searches for internal servers.

• Say, by sniffing traffic he discovers a mail server on the other

side of a firewall.
 Et cetera

• You can imagine how this should continue…

– The lesson, again, is think before you act.
 About the report

• The report will include

– A list of findings prioritized according to the risk (as
perceived by the pen-tester and taking into account the
scoping).
– A log of all what the pen-tester did (including compromises,
accidents & crashes,…).

• To prove that he broke in the different computers the

pen-tester can:
– Capture screenshots for workstations.
– Describe the explored network.
STUDENTS DISMISSED
 Closing remarks (for students)

What have students learned here?

1. Performing standard IG.

2. Detecting vulnerabilities and using exploits.
3. Pen-testing tactics:
• Pivoting.
• Think before taking an action.
4. Network security
• The three-legged DMZ configuration.
 Clossing remarks (for teachers)
What have students learned here?

1. By reading the module logs teacher can check whether

students used the right tools with the correct parameters.

2. Test the students’ ability to plan, e.g., did they perform

unnecessary actions.
– Let’s say we change firewall rules and network topology to another
(standard) scenario: can they guess the network topology and figure
out how to make the pentest?

3. Did they understand the essence of pivoting?

4. By understanding the students’ performance, the teacher can

identify their weaknesses as pentesters and plan new
exercises to work on these.
 So why is our learning kit necessary?

• But all this cannot be done in a small virtualization

lab/ farm! : Hardly!

• With our tool all students can do this at once!

– we require only one computer per student.
– in case the network / computers hangs, the simulation
environment can be easily reset.

• And there’s the problem of configuring new

scenarios.
– larger or more with more complex topologies.

• Last: students must be evaluated.

– success, performance, stealth, quality of reports.
QUESTIONS?
SIMULATING COMPUTER
ATTACKS - REALISTICALLY
 The tool is aimed at simulating network attacks

• A preliminary version is available, it runs on Win XP and Vista.

– Tradeoffs can be made to simulate 2k to 13K.

• It contains two main components:

– A trimmed & modified version of Core Impact Pen-testing
Framework.
which thinks it is connected to a real
network when actually it connects to:
– A network attack simulator that we call Core Insight.

Miranda, Orlicki, Sarraute, “Simulation of Computer Network Attacks.” AST`07

 It can simulate a wide variety of scenarios

• Simulated scenarios are composed of computers & network

devices.
– Each system can run processes and connect to other
devices.
– We simulate Windows workstations and servers, many Unix
systems, routers and firewalls.
– Each is configured independently according to the triple:
• OS, services, file system.
• Network connections to hubs, switches, dial-up connections.
• Vulnerabilities.

• In the current version, scenarios are generated by scripts.

 Simulation of sockets and syscalls
• Agents are controlled by the pen-tester, e.g., from Impact.
– There’s always a local agent + agents for compromised computers.

• Sockets are simulated. We support TCP and UDP sockets.

– We sort of wrapped a real BSD socket to do this.
– Raw packets and other more granular network operations are not
supported.

• Each (simulated) host is represented by a simulated syscall

server.
– So that the “controlling agent” will issue a syscall for the simulated host
(through thea standard interface), and receive its answer.
– Each system has at least a simulated thread (real thread for the host).
– Most of the syscalls are supported (but some are not).
 Simulation tradeoffs and the reality factor

• File systems simulation is supported.

– But to boost efficiency all systems are configured with a template file system by
default.
– When the template is modified, a separate file system is simulated.

• Some actions are only emulated (not simulated).

– For example, exploits are emulated: when an agent is commanded to send an
exploit, our simulator will look up for the outcome in a DB, e.g.,
ApacheChunkEncodingExploit yields compromise if the system should be
vulnerable.

• All actions have a probabilistic nature,

– E.g., ApacheChunkEncodingExploit against a RedHat 7 will work with %80
probability, crash the system with %5 and do nothing with %15.
– We have pre-computed probability distributions according to experimentation
with our model.
– The simulation kit has a DB that stores this information.
 pros & cons
PROS CONS
• This system provides a very good • There is a divergence
tradeoff between the simulation level between reality and
required to produce a pentest and the
hardware and setup requirements. simulation, e.g.,
– It requires less hardware than other – raw packet is not
solutions, e.g., can simulate more implemented.
systems with the same hardware.
– Setup of scenarios is much easier, as
well. • The current version
lacks support for client-
• The “module log” contains a list of all side, wireless, voip,
the modules executed and its results. scada, …
– this provides a tool for grading students.

• Support for hypothetic vulnerabilities.

QUESTIONS?
MORE LESSONS

The first lesson should be a glimpse of the methodology, and the

learning kit’s capabilities.
 The teaching program

• We want to complete this first lesson into a complete course.

• New lessons go in two directions

– Scenarios get more complicated (with network segmentation and more
firewalls, IDS detection that closes connections under noisy attacker
activity, …).
– No more abstract tools, as students must understand tools’ to use or
modify them.

• Plus
– Teaching exploit writing & engineering.
– Crypto protocols (e.g., I gained control of a box and can sign packets,
what can I do?).
– New attack vectors are explained (client-side, application, wireless, etc).
Going after the workstation first
 Other exercises are performed outside the kit

• Exploits
– Writing your own exploits (binary, application,…)
– Testing exploits: exploits should be reliable.
• Students can use a small virtualized farm with a handful of OS configurations
to this end.

• Getting to know the tools

– Sometimes in a real pen-test one needs non-standard tools, or modifying
known tools.

• Hybrid support
– Future versions of our kit will support hybrid simulation were simulated
computers are connected to real computers.
– An immediate application of this is webapps pen-testing with a real
webserver connected to a simulated network.
QUESTIONS?
OTHER APPLICATIONS OF THE
NETWORK ATTACK
SIMULATOR
 A research tool 1/2

1. Research automatic attack planning.

– Using the kit one could device planning algorithms that
execute pen-tests against input scenarios and optimize
them for… performance, stealth, etc.

2. Iterative risk assessment:

– Let’s say we make a pen-test against a real scenario using
Impact and succeed.
– “Import” all the discovered information to the simulator.
– Loop until you feel safe:
• Modify the simulated network.
• Attack it.
 A research tool 2/2

3. Generating logs (e.g., for training log analysis & IDSs).

1. Say, by executing several attacks against a given scenario.

4. Montecarlo analysis of different scenarios against a given

attack (or attack family).
1. We fix the scenario and vary on the attacks.
 ANY QUESTIONS ?

IN CLOSING MANY THANKS

E-mail:
ariel.waissbein/\at/\coresecurity.com
 Bibliography
• Multics Security Evaluation: Vulnerability Analysis, Karger, Schell (Air Force Electronic Systems
Division, 1974) http://csrc.nist.gov/publications/history/karg74.pdf

• The Protection of Information in Computer Systems, Saltzer and Schroeder (1975).

http://www.cs.virginia.edu/~evans/cs551/saltzer/

• Computer Security: the Achiles heel of the electronic air force, Schell (1979).
http://www.airpower.maxwell.af.mil/airchronicles/aureview/1979/jan-feb/schell.html

• Trusted Computer System Evaluation Criteria (aka, The orange book), National Security Institute -
5200.28-STD (1985). http://nsi.org/Library/Compsec/orangebo.txt

• The Internet Worm Incident, G. Spafford (1989). Technical Report CSD-TR-933, 1989.
http://homes.cerias.purdue.edu/~spaf/tech-reps/933.pdf

• Improving the security of your site by breaking into it (1993) - Dan Farmer, Wietse Venema.
http://www.porcupine.org/satan/admin-guide-to-cracking.html

• Syscall proxying, Max Caceres (2002). Black Hat Briefings 2002, Las Vegas.

• Modern Intrusion Practices, Gera Richarte (2003). Black Hat Briefings 2003, Las Vegas.

• Building Computer Network Attacks, A. Futoransky, L. Notarfrancesco, G. Richarte and C. Sarraute

(2003). http://www.corest.com/files/attachments/Futoransky_Notarfrancesco_Richarte_Sarraute_NetworkAttacks_2003.pdf
 Bibliography (2)
• Simulating computer network attacks, F. Miranda, J. Orlicki, C. Sarraute. AST `07 in
JAIIO `07, Mar del Plata, Bs. As., Argentina.
• Zombie 2.0, D. Tiscornia and F. Russ. HACK.LU 2007.
• Nmap, Fyodor. http://nmap.org. 1997.
• OS fingerprinting by neural networks, J. Burroni and C. Sarraute, PacSec 2004.
• Beyond Stack Smashing: Recent Advances in Exploiting Buffer Overruns, Jonathan
Pincus and Brandon Baker. IEEE Security & Privacy, July/Aug., pp. 20−27. 2004.