Read More - Dennis Technology Labs

Read More - Dennis Technology Labs
Network Threat Detection
DECEMBER 2015
Dennis Technology Labs
www.DennisTechnologyLabs.com
Follow @DennisTechLabs on Twitter.com
This report aims to compare the effectiveness of a
selection of threat detection appliances.
The products were exposed to internet threats
that were live during the test period, as well as
crafted exploit-based attacks.
These exposures were carried out in a realistic
way, closely reflecting a customer’s experience.
These results reflect what would have happened if
a user was using one of the products and visited an
infected website or was attacked in a more
targeted way such as when receiving infected
documents and customized links to infected
websites.
This is purely a detection test, not a protection
test. The threat detection appliances were
deployed in ‘tap mode’ so that they could monitor
network traffic but would not be able to block
threats.
No additional anti-malware products were
installed on the endpoints.
EXECUTIVE SUMMARY
Products tested
Product
Detection accuracy
Legitimate accuracy
Symantec Advanced Threat Protection
100%
100%
Palo Alto Networks PA200
90%
97%
Cisco Snort
72%
100%
Fortinet FortiGate60D
69%
100%
Product names
The products tested in this report were the latest
versions available from each vendor on the date
that the test started.
Specific ‘build numbers’ are available for those who
wish to ascertain the exact versions that were
used for testing.
Additionally, many products offer different
protection modules, including URL filtering, antimalware detection and cloud-based sandboxes for
malware execution. These were all enabled.
These are listed in Appendix C: Product versions
on page 13.

Network threat detection systems miss even well-known threats.
The majority of the threats used in this test were live web-based threats that were attacking users on the
internet at the same time as the test was run. With the notable exception of Symantec’s product, most
detection appliances failed to spot all of the threats, with two recognizing less than 75 per cent.

The appliances were accurate when handling legitimate software
None of the products blocked the installation of legitimate software and only one rated a moderately
popular application as being suspicious.

Which was the best product?
The most accurate appliance was Symantec Advanced Threat Protection. This was followed by a good
performance by Palo Alto Networks’ PA200.
Simon Edwards, Dennis Technology Labs, 18th Dec 2015
Network Threat Detection, December 2015
Page 2 of 14
CONTENTS
Executive summary .................................................................................................................................................................... 2
Contents ....................................................................................................................................................................................... 3
1. Detection Scores ................................................................................................................................................................... 4
2. Legitimate Software Ratings ................................................................................................................................................ 5
3. The Tests ................................................................................................................................................................................. 8
4. Test Details ............................................................................................................................................................................. 8
5. Conclusions ........................................................................................................................................................................... 11
Appendix A: Terms Used ....................................................................................................................................................... 12
Appendix B: FAQs.................................................................................................................................................................... 13
Appendix C: Product versions .............................................................................................................................................. 14
Document version 1. 0. Written 18th Dec 2015.
Network Threat Detection, December 2015
Page 3 of 14
1. DETECTION SCORES
The total accuracy ratings provide a way to judge
how effectively the security programs work by
looking at a single graph.
The results below take into account how
accurately the programs treated threats and
handled legitimate software.
Anti-malware software should not just detect
threats. It should allow legitimate software to run
unhindered as well.
Detection Scores
100
90
80
30
20
10
Cisco Snort
40
Palo Alto Networks
PA200
50
Symantec Advanced
Threat Protection
60
Fortinet FortiGate60D
70
0
Every time a product detected a threat it was awarded one point. It gained no points for failing to register
an attack in its logs.
It is important to view the detection scores
alongside the legitimate software ratings on page 5.
The ideal product would detect all threats and not
generate false alerts when encountering legitimate
applications.
When a product fails to detect a threat the
network administrators will be less able to react
effectively to a successful attack. When it warns
against legitimate software then it generates a ‘false
positive’ result and network administrators risk
wasting time on unnecessary alerts or, worse, will
fail to notice important accurate alerts that are
lost in a mass of false alerts.
See 2. Legitimate Software Ratings on page 5 for
detailed results and an explanation on how the
false positive ratings are calculated.
DETECTION SCORES
Product
Detected Scores
Symantec Advanced Threat Protection
100
Palo Alto Networks PA200
90
Cisco Snort
72
Fortinet FortiGate60D
69
Network Threat Detection, December 2015
Page 4 of 14
2. LEGITIMATE SOFTWARE RATINGS
The legitimate software accuracy ratings provide a
way to judge how effectively the security programs
handle non-malicious software by looking at a
single graph.
Anti-malware software should allow legitimate
software to run unhindered. These results take
into account the level of any interaction that the
product demands of the user, as well as the
prevalence of the legitimate program.
To understand how we calculate these ratings see
2.3 Accuracy ratings on page 7.
Legitimate Software Ratings
134
120.6
107.2
67
40.2
26.8
13.4
Cisco Snort
53.6
Palo Alto Networks
PA200
Fortinet FortiGate60D
80.4
Symantec Advanced
Threat Protection
93.8
0
When a product misclassified a popular program it faced a stronger penalty than if the file was more
obscure.
LEGITIMATE SOFTWARE RATINGS
Product
Accuracy Rating
Cisco Snort
134
Fortinet FortiGate60D
134
Symantec Advanced Threat Protection
134
Palo Alto Networks PA200
129.5
Network Threat Detection, December 2015
Page 5 of 14
 2.1 Interaction ratings
A threat detection product needs to be able to
detect threats, while not generating false alerts
when encountering legitimate software. When
legitimate software is misclassified as malware a
false positive is generated.
A product gains top marks if it allows legitimate
software to install without requiring the user to
answer questions or otherwise interact. It loses
points the more interaction is required and the
less accurately it behaves.
Some security products will ask the user questions
when they encounter software and are not certain
if it is either fully legitimate or definitely malware.
If a product actually generates a genuine false
positive (e.g. “software is malicious”) it is penalized
heavily.
When measuring how effective each product is we
take into account all of the likely outcomes:
whether the product allows, blocks or asks
different types of questions. In each case a score is
allocated.
The results grid below shows the most likely
possibilities, along with some outcomes that could
only happen if a product was not working properly
(e.g. A5 – Object is safe but is blocked
automatically).
Classification
Interaction
None
Click to allow Click to allow/block Click to block None
(allowed) (default allow) (no recommendation) (default block) (blocked)
Object is safe
2
1.5
1
X
X
Object is unknown
2
1
0.5
0
-0.5
Object is not classified
2
0.5
0
-0.5
-1
Object is suspicious
0.5
0
-0.5
-1
-1.5
Object is unwanted
0
-0.5
-1
-1.5
-2
Object is malicious
X
X
X
-2
-2
1
2
3
4
5
A
B
C
D
E
F
Top marks to products that are accurate; those that ask too many questions or are overly suspicious are
penalized.
LEGITIMATE SOFTWARE INCIDENTS
Product
Classification
Total
Palo Alto Networks PA200
Object is suspicious
1
Network Threat Detection, December 2015
Page 6 of 14
 2.2 Prevalence ratings
The prevalence of each piece of software is
significant. If a security product interferes with
common applications then the situation is more
serious than if it does so with rare ones. That said,
it is usually expected that anti-malware programs
should not interfere with any legitimate software.
The programs selected for the legitimate software
testing were organized into five groups:
Very High Impact; High Impact; Medium Impact;
Low Impact; and Very Low Impact.
The table below shows the relative importance of
each group expressed as a numerical value. A Very
High Impact application is ranked as being five
times more significant than a Very Low Impact
program.
LEGITIMATE SOFTWARE
PREVALENCE RATING MODIFIERS
Impact category
Very High Impact
High Impact
Medium Impact
Low Impact
Very Low Impact
Rating modifier
5
4
3
2
1
These categories were attributed to software
programs based on their individual weekly
download numbers as reported by third-party
download sites including Download.com at the
time of testing.
Files were downloaded from their original sources,
excluding third-party download sites, such as
Download.com, wherever possible. This was to
reduce the chances that the software had been
altered in any way, perhaps having potentially
unwanted add-ons included with the installer.
The presence of potentially unwanted add-ons
transforms the legitimate software into a product
that could be blocked or altered justifiably by antimalware software. As such they are not suitable
for this legitimate software test.
The ranges for these categories, in terms of
weekly downloads, are recorded in the table
Legitimate Software Prevalence Categories.
Network Threat Detection, December 2015
LEGITIMATE SOFTWARE
PREVALENCE CATEGORIES
Impact category
Prevalence
Very High Impact
High Impact
Medium Impact
Low Impact
Very Low Impact
>20,000
1,000 – 20,000
100 – 999
25 – 99
< 25
 2.3 Accuracy ratings
The legitimate software accuracy ratings are
calculated by multiplying together the interaction
and prevalence ratings.
accuracy rating = number of programs x
(interaction rating x prevalence rating)
For example, if a product allows 10 legitimate,
Medium Impact programs to install without any
interference then its rating would be calculated
like this:
accuracy rating = 10 x (2 x 3)
= 60
This formula creates the impact-weighted accuracy
ratings used in the graph 2. Legitimate Software
Ratings on page 5.
 2.4 Distribution of impact categories
Products that scored highest were the most
accurate when handling the legitimate applications
used in the test.
The best theoretical score possible is 1,000, while
the worst would be -1,000 (assuming that all
applications were classified as Very High Impact).
In fact the distribution of applications in the impact
categories was not restricted only to Very High
Impact. The table below shows the true
distribution:
LEGITIMATE SOFTWARE
CATEGORY FREQUENCY
Prevalence Rating
Frequency
Very High Impact
2
High Impact
10
Medium Impact
3
Low Impact
3
Very Low Impact
2
Page 7 of 14
3. THE TESTS
 3.1 The threats
Providing a realistic user experience was important
in order to illustrate what really happens when a
user encounters a threat on the internet.
For example, in these tests web-based malware
was accessed by visiting an original, infected
website using a web browser, and not downloaded
from a CD or internal test website.
All target systems were fully exposed to the
threats. This means that any exploit code was
allowed to run, as were other malicious files, They
were run and permitted to perform exactly as they
were designed to, subject to checks made by the
installed security software.
A minimum time period of five minutes was
provided to allow the malware an opportunity to
act.
 3.2 Test rounds
Tests were conducted in rounds. Each round
recorded the exposure of every product to a
specific threat. For example, in ‘round one’ each of
the products was exposed to the same malicious
website.
At the end of each round the test systems were
completely reset to remove any possible trace of
malware before the next test began.
 3.3 Monitoring
Close logging of the target systems was necessary
to gauge the relative successes of the malware.
This included recording activity such as network
traffic, the creation of files and processes and
changes made to important files.
4. TEST DETAILS
 4.1 Product configuration
The threat detection systems were configured as
network taps, which means that they were given
the opportunity to monitor network traffic and to
detect threats that passed through that traffic.
The main advantage of ‘tap mode’ is that even if
the system fails the network is still able to operate.
The usual alternative, ‘inline mode’ allows products
the chance to block threats as well as detect them.
The downside is that if the product fails the
normal network traffic could be disrupted.
A selection of legitimate but vulnerable software
was pre-installed on the target systems deployed
on the network behind the detection systems. The
software installed on these target systems posed
security risks, as they contained known security
issues. They included versions of Adobe Flash
Player, Adobe Reader and Java.
Due to the dynamic nature of the tests, which
were carried out in real-time with live malicious
websites and customized attacks, the products'
update systems were allowed to run automatically
and were also run manually before each test round
was carried out.
Network Threat Detection, December 2015
The products were also allowed to 'call home'
should they be programmed to query databases in
real-time. Some products might automatically
upgrade themselves during the test. At any given
time of testing, the very latest version of each
firmware was used.
Each target systems was a physical PC, not a
virtual machine, and was connected to the internet
via its own virtual network (VLAN) to avoid crossinfection of malware.
 4.2 Threat selection
The malicious web links (URLs) and other samples
used in the tests were not provided by any antimalware vendor.
Live URLs were picked from lists generated by
Dennis Technology Labs’ own malicious site
detection system, which uses popular search
engine keywords submitted to Google. It analyses
sites that are returned in the search results from a
number of search engines and adds them to a
database of malicious websites.
In all cases, a control system (Verification Target
System - VTS) was used to confirm that the URLs
linked to actively malicious sites.
Page 8 of 14
Custom attacks were generated using Metasploit,
with minimal changes to the default settings. When
these threats were web-based, they were
deployed as part of apparently legitimate websites,
effectively producing man-in-the middle attacks at
the external network segment. In this way targets
and threat detection systems have a chance to
detect internet-based attacks from real websites
using custom threats.
Malicious URLs and files were not shared with any
vendors during the testing process.
 4.3 Test stages
There were two main stages in each individual test:
1.
2.
Introduction
Observation
During the Introduction stage, the target system
was exposed to a threat. Before the threat was
introduced, a snapshot was taken of the system.
This created a list of Registry entries and files on
the hard disk. The threat was then introduced.
Immediately after the system’s exposure to the
threat, the Observation stage is reached. During this
time, which typically lasted at least 10 minutes, the
tester monitored the system both visually and
using a range of third-party tools.
In the event that hostile activity to other internet
users was observed, such as when spam was being
sent by the target, this stage was cut short.
The Observation stage concluded with another
system snapshot. This ‘exposed’ snapshot was
compared to the original ‘clean’ snapshot and a
report generated. The system was then rebooted.
All log files, including the snapshot reports and the
product’s own log files, were recovered from the
target.
In some cases the target may become so damaged
that log recovery is considered impractical. The
target was then reset to a clean state, ready for
the next test.
Logs from the threat detection systems were
collected both in real-time and at the end of the
test. Testers also recorded any relevant activity
visible to the user of the target systems, such as
messages sent to the browser regarding URL
blocking.
Network Threat Detection, December 2015
 4.4 Threat introduction
Malicious websites were visited in real-time using
the web browser. This risky behavior was
conducted using live internet connections. URLs
were typed manually into the browser.
Web-hosted malware often changes over time.
Visiting the same site over a short period of time
can expose systems to what appear to be a range
of threats (although it may be the same threat,
slightly altered to avoid detection).
Also, many infected sites will only attack a
particular IP address once, which makes it hard to
test more than one product against the same
threat.
In order to improve the chances that each target
system received the same experience from a
malicious web server, we used a web replay
system.
When the verification target systems visited a
malicious site, the page’s content, including
malicious code, was downloaded, stored and
loaded into the replay system. When each target
system subsequently visited the site, it received
exactly the same content.
The network configurations were set to allow all
security products unfettered access to the internet
throughout the test, regardless of the web replay
systems.
 4.5 Observation and intervention
Throughout each test, the target and threat
detection systems were observed both manually
and in real-time. This enabled the testers to take
comprehensive notes about the system’s perceived
behavior, as well as to compare visual alerts with
the products’ log entries.
At certain stages the tester was required to act as
a regular user. To achieve consistency, the tester
followed a policy for handling certain situations,
including dealing with pop-ups displayed by
products or the operating system, system crashes,
invitations by malware to perform tasks and so on.
This user behavior policy included the following
directives:
1.
Act naively. Allow the threat a good
chance to introduce itself to the target by
clicking OK to malicious prompts, for
example.
Page 9 of 14
2.
3.
4.
5.
6.
Don’t be too stubborn in retrying blocked
downloads. If a product warns against
visiting a site, don’t take further measures
to visit that site.
Where malware is downloaded as a Zip
file, or similar, extract it to the Desktop
then attempt to run it. If the archive is
protected by a password, and that
password is known to you (e.g. it was
included in the body of the original
malicious email), use it.
Always click the default option. This
applies to security product pop-ups,
operating system prompts (including
Windows firewall) and malware
invitations to act.
If there is no default option, wait. Give
the prompt 20 seconds to choose a
course of action automatically.
If no action is taken automatically, choose
the first option. Where options are listed
vertically, choose the top one. Where
options are listed horizontally, choose the
left-hand one.
 4.6 Automatic monitoring
Logs were generated using third-party applications,
as well as by the security products themselves.
Manual observation of the target system
throughout its exposure to malware (and
legitimate applications) provided more information
about the security products’ behavior.
Monitoring was performed directly on the target
system and on the network.
Network Threat Detection, December 2015
Client-side logging
A combination of Process Explorer, Process
Monitor, TcpView and Wireshark were used to
monitor the target systems. Regshot was used
between each testing stage to record a system
snapshot.
A number of Dennis Technology Labs-created
scripts were also used to provide additional
system information. Each product was able to
generate some level of logging itself.
Process Explorer and TcpView were run
throughout the tests, providing a visual cue to the
tester about possible malicious activity on the
system. In addition, Wireshark’s real-time output,
and the display from the web proxy (see Network
logging, below), indicated specific network activity
such as secondary downloads.
Process Monitor also provided valuable
information to help reconstruct malicious
incidents.
Network logging
All target systems were connected to a live
internet connection, which incorporated a
transparent web proxy and a network monitoring
system. All traffic to and from the internet had to
pass through this system.
An HTTP replay system ensured that all target
systems received the same malware as each other.
It was configured to allow access to the internet
so that threat detection systems could download
updates and communicate with any available ‘in the
cloud’ servers.
Page 10 of 14
5. CONCLUSIONS
 Where are the threats?
The threats used in this test were genuine, real-life
threats that were infecting victims globally at the
time that we tested the products.
The types of infected or malicious sites were
varied, which demonstrates that effective threat
detection is essential for businesses that wish to
understand the nature of the internet threats
facing their networks.
Most threats installed automatically when a user
visited the infected webpage. Such infections were
often invisible to a casual observer and so network
administrators should not expect users to be able
to manually alert them about attacks.
 Alerts, false alerts and silence
All products succeeded in avoiding any significant
level of false positive alerts.
In one minor incident the Palo Alto device claimed
that one file, which was moderately popular on the
internet at the time, was “suspicious”. Other than
that the products were accurate when handling
legitimate URLs and software.
This level of accuracy is not reflected in the threat
detection scores.
Network Threat Detection, December 2015
While Symantec’s appliance detected all of the
threats and Palo Alto’s alerted on 90 per cent of
the same, both Cisco Snort and Fortinet detected
less than 75 per cent.
 Next stop, protection
Clearly if a network-based threat detection
systems are not universally capable of detecting
threats then other, additional measures are
needed to discover how attackers are attempting
to compromise business networks.
Running detection systems inline, as opposed to in
tap mode, will help provide some protection but,
in the case of Cisco Snort and Fortinet FortiGate
60D, even this approach would not be sufficient to
prevent successful public web-based attacks, let
alone more targeted attacks that are customized
to breach a specific organization.
It is highly likely, based on previous research by
Dennis Technology Labs, that deploying endpoint
anti-malware solutions, updating vulnerable
software and blocking infected emails will improve
an organization’s level of protection.
For more details about the protection levels
provided by such measures please visit
www.DennistechnologyLabs.com.
Page 11 of 14
APPENDIX A: TERMS USED
Compromised
Malware continues to run on an infected system, even after an on-demand scan.
False Positive
A legitimate application was incorrectly classified as being malicious.
Introduction
Test stage where a target system is exposed to a threat.
Observation
Test stage during which malware may affect the target.
Prompt
Questions asked by software, including malware, security products and the operating
system. With security products, prompts usually appear in the form of pop-up windows.
Some prompts don’t ask questions but provide alerts. When these appear and
disappear without a user’s interaction, they are called ‘toasters’.
Remediation
Test stage that measures a product’s abilities to remove any installed threat.
Round
Test series of multiple products, exposing each target to the same threat.
Snapshot
Record of a target’s file system and Registry contents.
Target
Test system exposed to threats in order to monitor the behavior of security products.
Threat
A program or other measure designed to subvert a system.
Update
Code provided by a vendor to keep its software up to date. This includes virus
definitions, engine updates and operating system patches.
Network Threat Detection, December 2015
Page 12 of 14
APPENDIX B: FAQS








This test was sponsored by Symantec.
The test rounds were conducted between 12th October 2015 and 3rd December 2015 using the most up
to date versions of the appliances available on any given day.
All products were able to communicate with their back-end systems over the internet.
The products selected for this test were chosen by Symantec under close consultation with Dennis
Technology Labs.
Samples were located and verified by Dennis Technology Labs.
Products were exposed to live threats within 24 hours of the same threats being verified. In practice there
was only a delay of up to three to four hours.
Details of the samples, including their URLs and code, were provided to Symantec only after the test was
complete.
The sample set comprised 75 actively-malicious URLs, 25 custom threats and 20 legitimate applications.
Do participating vendors know what samples are used, before or during the test?
No. We don’t even know what threats will be used until the test starts. Each day we find new ones, so it is
impossible for us to give this information before the test starts. Neither do we disclose this information until
the test has concluded.
Do you share samples with the vendors?
The sponsor vendor is able to download samples from us after the test is complete.
Other vendors may request a small subset of the threats that compromised their products in order for them
to verify our results and further understand our methodology. The same applies to client-side logs, including
the network capture files. There is a small administration fee for the provision of this service.
What is a sample?
In our tests a sample is not simply a set of malicious executable files that runs on the system. A sample is an
entire replay archive that enables researchers to replicate the incident, even if the original infected website is
no longer available. This means that it is possible to reproduce the attack and to determine which layer of
protection is was able to bypass. Replaying the attack should, in most cases, produce the relevant executable
files. If not, these are usually available in the client-side network capture (pcap) file.
Network Threat Detection, December 2015
Page 13 of 14
APPENDIX C: PRODUCT VERSIONS
A product’s update mechanism may upgrade the software to a new version automatically so the version used
at the start of the test may be different to that used at the end.
Vendor
Product
Build
Cisco
Snort
2.9.7.6
Fortinet
FortiGate 60D
Fortigate-60D v5.2.4 , build 0688,150722 (GA)
Palo Alto
Networks
PA200
Symantec
Advanced Threat Protection
PAN OS :6.0.9; App-version: 537-2965
2.0.0-58
WHILE EVERY EFFORT IS MADE TO ENSURE THE ACCURACY OF THE INFORMATION PUBLISHED IN
THIS DOCUMENT, NO GUARANTEE IS EXPRESSED OR IMPLIED AND DENNIS PUBLISHING LTD DOES
NOT ACCEPT LIABILITY FOR ANY LOSS OR DAMAGE THAT MAY ARISE FROM ANY ERRORS OR
OMISSIONS.
Network Threat Detection, December 2015
Page 14 of 14
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement