In-Depth
Barbarians at the Gate
Intrusion detection systems are becoming increasingly important in network security. Here’s a primer on what they do and how they work—and an evaluation of four popular sentries.
When discussing network security, many IT professionals associate security
technologies with network access control mechanisms such as firewalls,
proxy servers and router access control lists. They’re designed to restrict
the flow of traffic into and out of corporate networks to reduce the risk
of compromise by malicious individuals on the Internet. However, even
the best filtering devices do very little to protect a network if Internet-accessible
devices are vulnerable to attacks. Malicious individuals can use open
communications channels (such as HTTP, e-mail and DNS) to compromise a
system on a network remotely, and use that compromised system as a launching
pad to attack and compromise other “trusted” systems on the network.
Many of today’s IT infrastructures have become so large, complex and
geographically dispersed that IT managers may not learn about the defacement
of their Web page in a foreign data center for several days. In the case
of one organization, an IT manager learned of a system compromise after
a user notified the corporate public relations office. Public relations
notified a top corporate officer, who relayed the information to the manager.
This isn’t how most of us would like to learn about such occurrences.
The good news is that intrusion detection systems (IDS) can fill such
security gaps by providing a way to monitor networks for the presence
of active exploits and misuse. This article provides a brief overview
of network-based intrusion detection that can help you evaluate your current
security infrastructure.
What is IDS?
Intrusion detection is a security process designed to monitor and
analyze system or network events to detect possible misuse or unauthorized
access. Broadly defined, IDS are technologies intended to support the
intrusion detection process. They can be categorized into network-based
(NIDS) and host-based systems (HIDS). Within these two categories, a number
of sub-categories have emerged: embedded IDS, honeypots, file integrity
checkers and even anti-virus scanners. New concepts and terms are evolving
in this field as the technology progresses and vendors attempt to introduce
new functionality and differentiate their products in this growing space.
Companies are beginning to build upon concepts such as protocol anomaly
detection and distributed IDS event correlation; but the majority of the
systems currently in use are classified as pattern-matching systems.
IDS is similar to a motion detector installed in your house that’s connected
to a central alarm system. When locks fail to keep intruders outside,
the sensor will detect motion and trigger an alert. Of course, it’s important
to locate the sensor properly in your house and activate it. After all,
a sensor in your garage won’t help against intruders breaking in through
your kitchen window.
Product
Information |
Dragon
Starts at $3,000 for 10 MB of network bandwidth,
or $750 per host.
Enterasys Networks
Portsmouth, New Hampshire
603-501-5500
www.enterasys.com/ids/
ISS RealSecure
$8,995 per server for version 7.0.
Internet Security Systems
Atlanta, Georgia
404-236-2600
www.iss.net
NetProwler
1-50 nodes, $2,995; 1-150 nodes, $5,995
Symantec Corp.
Cupertino, California
408-253-9600 http://enterprisesecurity.symantec.com/
Snort with ACID Console
Free (open source)
www.snort.org
|
|
|
Network Intrusion Detection Systems
NIDS monitor network traffic and report on specific events on the
network. NIDS are usually based on two principles: pattern-matching and
statistical or algorithmic anomaly detection. Pattern-matching deals with
identifying a specific sequence or attribute within a packet, such as
matching source or destination IP address or certain characters within
packet headers and payloads. Statistical or algorithmic anomaly detection
identifies whether the traffic meets certain benchmarks, such as detecting
the number of SYN packets sent per minute or a number of connections attempted
in a given period.
In NIDS terms, the patterns are usually called “signatures,” and statistical
or algorithmic benchmarks are called “rules.” For example, the Code Red
worm can be identified within the network traffic as a packet addressed
to a destination HTTP port 80 and containing the following pattern (or
signature) within its payload (simplified for publication):
GET /default.ida? XXXXXXXXXXXXXXXXX XXXXXXXXXXXXXX%u9090%u6858%
Similarly, unauthorized login attempts or brute force attacks can be
detected via reporting on a statistical or algorithmic anomaly (NIDS rule)
of 10 to 20 unsuccessful FTP login requests per minute. Modern NIDS can
be programmed to report almost any kind of hypothetical anomaly.
It’s important to mention that many NIDS products (including those reviewed
in this article) allow network administrators to create signatures and
rules. Therefore, NIDS can be used not only to detect active attacks,
but also for network troubleshooting purposes or to fish for any packets
that could be interesting to networking personnel. For example, we’ve
found it useful to watch for the use of “public” and “private” community
strings in SNMP packets, as our corporate policy dictates that all SNMP
community strings must be changed from the default. We’re able to use
NIDS to identify systems that don’t comply with this rule.
Of course, the ability of an IDS system to do its job implies that the
IDS sensor is capable of “sniffing” all traffic to and from monitored
systems and matching each packet with the database of malicious signatures.
This fact necessitates the placement of the NIDS on network segments where
all traffic to be monitored can be captured. The use of mirrored or span
ports on switches or placement hubs on the uplink ports of switches or
router interfaces can usually get the job done. In redundant network architecture,
each redundant link must have its own NIDS. Avoid creating single points
of failure on the network when designing an intrusion detection solution.
The majority of studies show that most malicious attacks occur on the
internal network, and organizations should seriously consider implementing
IDS not only in the DMZ or other externally accessible segments, but also
on the internal network. Threats of unauthorized access from within the
internal network are also high given the growing number of third parties
(consultants, vendors and employees) having access to corporate networks.
The biggest challenge in monitoring the internal network for malicious
activity is that such activity may come from authorized access—users may
not have to use hacking techniques reportable by the IDS. For that reason,
more and more organizations choose to use host-based intrusion detection
systems (HIDS) in conjunction with the NIDS.
Host-Based Intrusion
Host-based intrusion detection systems are designed to detect host-based
events, such as using the Windows 2000 “Run As” command or gaining legitimate
access to a confidential file, which may signify unauthorized access if
issued during off hours. Host-based IDS mechanisms range from basic system
auditing to advanced detection methods and have proved to be most efficient
in tracking potential system misuse from users on the internal network.
In many cases, IDS vendors offer both network and host-based systems,
consolidating the alerts into a common reporting console. While host-based
intrusion detection is an important component of an organization’s security
strategy, this article focuses on NIDS, the network-based approach.
Knowledge,
Process and Technology |
While IDS is an important part of the organizational
defense posture, it’s only effective when coupled with
knowledge and established processes.
By knowledge, we mean your in-depth understanding of
network protocols, awareness of new attacks, and understanding
of your networks (in other words, how layer 2 and 3
traffic is flowing through the network to and from specific
servers). By process, we mean diligent application of
signature upgrade procedures, continuous monitoring,
ability to validate certain events, and your preparedness
to react properly to detected events.
IDS itself won’t change the security of your organization
overnight, but, when combined with knowledge and processes,
it will slowly but surely pinpoint current network deficiencies
and enable you to react in response to detected attacks.
|
|
|
IDS is Not a Cure-All
Intrusion detection systems aren’t the magic security bullet some
people perceive them to be. Encryption, designed to prevent unauthorized
disclosure of information, is particularly troublesome for intrusion detection
systems. Widely used network security protocols such as SSL and SSH, as
well as IPSec used in virtual private networks, preclude NIDS from inspecting
network traffic for comparison against attack signatures. Encrypted payloads
make these protocols excellent vehicles for an attacker to evade detection
by an IDS.
Another important limitation for NIDS is network speed. Hardware limitations
and inefficiencies in pattern-matching algorithms impose limitations on
the number of packets that can be processed for a given period. Once this
limit is reached, IDS usually starts “dropping” packets (ignoring the
traffic). Few NIDS available on the market today can work on gigabit networks.
To make it easier for the black hat community to attack infrastructures
protected with NIDS, a number of tools that circumvent the operation of
NIDS sensors have emerged. Some of these tools (such as Stick, which is
described later) are able to send an entire database of attack signatures
past a NIDS sensor and make the console light up like a Christmas tree
generating false alerts. Organizations must have a clear understanding
of the risks such tools represent to the NIDS’ operations and how to react
to these alerts. Usually the attackers try to evade detection by your
NIDS in order to execute real attacks and make it hard (if not impossible)
to detect them.
Despite what the vendors say, effective NIDS are a relatively high-maintenance
technology. They can’t merely be configured once on the network and left
to run by themselves. New signatures, network changes and new applications
all require IDS tweaking. To achieve maximum effectiveness, signatures
must be frequently updated, thresholds must be tuned to reduce false positives,
and administrators must monitor output to obtain an understanding of normal
traffic patterns for their environment. Additions and deletions of network
components and servers must be accompanied by necessary changes to the
NIDS. Failure to do so may result in being unable to pick up important
events related to the newly deployed production subnets. These activities
can be resource intensive if proper processes and administrative resources
aren’t in place. To alleviate the strain these requirements may impose,
a number of third-party managed security service offerings have sprung
up to handle outsourcing of these activities.
Today, NIDS can also be configured to manage the traffic proactively
on their own or by communicating with firewalls and other network components
in response to events. For example, ISS RealSecure, Snort and Dragon can
reset TCP sessions based on a specific trigger (by sending an RST packet
to the source). ISS RealSecure can also integrate with Microsoft ISA Server
and Checkpoint FireWall-1 to cause an automated configuration change based
on certain events. (We should note that if you’re familiar with Dragon,
then you know it’s a purely Unix tool. Why is it showing up in these pages?
Because these days most IDSs are Unix-based—and Dragon’s an excellent
example of the genre.)
5
Deployment Caveats |
- Deployment of IDSs need to account for VPN, IPsec
and other encryption usage. For example, if you allow
encrypted traffic to and from Web and application
servers, IDS may not be able to help.
- For ultimate defense, IDSs must cover redundant
and fail-over links. This may be an engineering challenge
(how do you properly match a signature that happens
to be split among several wires?) and budgetary pressure
due to extra hardware and software requirements.
- Make sure sensitive administrative information
is properly secured. In one case, we noticed a NIDS
capturing sensitive information, which could have
been used to gain access to corporate firewalls and
perform reconfiguration. The IDS was outsourced and
firewalls were managed in-house—not a separation of
duties we’d want to rely on! (Does your IDS capture
your telnet password?)
- How IDS is “tapped” into networks may significantly
affect the usability of this technology. Make sure
you consult your router and switch vendor prior to
any implementations. Some vendors have limitations,
which don’t allow for capturing of all the traffic
on a specific device.
- Last but not least, users and customers may need
to be notified about the presence of such technology
on your network. We believe, however, that in most
(but not all) cases, the presence of NIDS should remain
a secret and NIDS shouldn’t respond to any network
events. The fact that user traffic is monitored may
have legal implications and may have to be disclosed
to the user community via a high-level communication.
|
|
|
Can They Be Trusted?
Most industry experts agree that intrusion detection systems haven’t
reached the point where they can be trusted to respond to attacks by automatically
reconfiguring network equipment. The problem with automatic response to
network events (such as system reconfiguration) is that your IDS installation
becomes a potential threat. If attackers realize that you block port 80
to the source IP of a malicious packet, they can easily spoof legitimate
source IPs of your partners and clients in the malicious packets addressed
to your Web server, thus denying service to the legitimate users. However,
if you perceive a denial-of-service as having less impact on your environment
than unauthorized access, you may still choose to deploy automatic system
reconfiguration.
Another complication comes from the architecture of IDSs. In many cases,
enterprise deployments require communication between agents (traffic-sniffing
devices) and managers (devices that govern the distribution of policies
and how alerts are processed between the management stations and consoles).
For this, firewalls need to be configured for the “trusted” IP addresses
in the DMZ (or even worse, on the Internet) to be allowed access to the
network management station. If proper consideration isn’t given to such
configurations, NIDS deployment may lead to a lot of damage. In some cases,
you might need to create a separate management network to use additional
network interfaces on the agent machines.
To summarize, the following are the tasks generally performed by a network
IDS:
- Near real-time analysis of network traffic for known attack signatures.
- Near real-time detection and reporting of network traffic anomalies.
- Alerting administrative personnel when a potential attack is detected.
- Taking a pre-defined corrective action in response to an event.
- Sophisticated non-real time reporting and data analysis.
The Test Network
To perform our product evaluations, we set up a Win2K (SP2)-based
IIS 5.0 Web server in a test lab. For the sake of simplicity, the network
interface of the Web server was plugged into the same hub as the NIDS
agents, allowing for the NIDS to capture all traffic directed toward the
Web server. We implemented no firewalls to filter traffic to or from the
Web server. We performed testing on a LAN, as well as across the Internet.
The Attackers
We used the following tools to scan the Web server for services
and vulnerabilities, as well as execute attacks against the server and
the NIDS:
- Nmap port scanner. Nmap is a fast, efficient port scanner.
Nmap TCP stealth port scan, UDP scan and ping sweep were executed against
the Web server. www.insecure.org
- ISS Internet Scanner. This is one of the most popular port
and vulnerability scanners. Vulnerability scans, as well as actual denial-of-service
(DoS) and buffer overflow attacks were executed against the Web server.
www.iss.net
- Nessus vulnerabity scanner. Nessus is a popular open source
vulnerability scanner. Vulnerability scans, as well as actual DoS and
buffer overflow attacks were executed against the Web server. www.nessus.org.
- Arirang CGI scanner. Arirang is a fast, efficient scanner
designed to identify common gateway interface (CGI) vulnerabilities.
www.freebsd.org/ports/security.html.
- Stick. Stick is an IDS obfuscation tool. It’s designed to
send attack signatures from a Snort signature file against the target,
creating a large number of false alarms on the NIDS. Note: Because Stick
spoofs legitimate source IP addresses, it should never be used on networks
connected to the Internet. http://packetstorm.dnsi.info/distributed.
- Fragroute. Fragroute’s objective is the opposite of Stick:
It fragments outgoing malicious packets within the source of attack
in a way that’s supposed to make the target NIDS blind to these attacks.
www.monkey.org/~dugsong/fragroute.
The Contenders
We tested the functionality of the following intrusion detection
systems.
NetProwler 3.5.1
Symantec NetProwler (Figure 1) supports Windows 98, NT and Win2K.
It’s an old-time IDS system; the current version dates back to August
2000, which is probably attributable to the acquisition of Axent by Symantec.
NetProwler Commercial support is included in the base contract.
|
Figure 1. NetProwler gives a detailed breakdown
of a network attack. |
NetProwler consists of console, manager and agent modules. Agent and
manager work only on NT 4.0 with SP3 and later. Console works on Windows
9x, Win2K and NT. (It may work on XP but we haven’t tested this). NetProwler
has a signature synchronization feature that allows for an automatic download
of newer rules. The product allows users to configure various alerts as
well as execute various commands based on triggered signatures. Users
can push policies (containing signatures and rules) to remote NIDS sensors,
thus allowing for centralized remote administration.
ISS RealSecure 6.5
ISS RealSecure 6.5 supports Windows NT 4.0/2000, SunOS 2.5.1, Solaris
2.6, 7, 8, HP-UX 11, IBM AIX 4.3, Nokia Appliance IPSO, and Linux RH 7.1.
Both NIDS and HIDS agents can be supported from the same console. You
can customize policies and select specific signatures. Most of the signatures
for the known attacks are proprietary and must be downloaded from the
vendor via a remote update. ISS Real Secure also allows users to define
signatures via a simple API interface. Commercial support is included
in the base contract, and managed service offerings are available.
Installation of RealSecure 6.5 on Win2K requires MSDE2000, the RealSecure
Console (Figure 2), Event Collector, and sensor (network or host/server).
All communications between the agents and the event collectors are encrypted.
Users must provide a license key during the installation.
|
Figure 2. ISS’s RealSecure allows users to select
which attacks should be reported and how the reporting should occur.
(Click image to view larger version.) |
Among the features provided by RealSecure are automated signature updates,
signature selection, signature creation (simple API), manipulation of
network traffic (connection reset), plug-ins for other products (ISA server,
Checkpoint FW-1), session recording and playback, SNMP traps, e-mail alerts,
remote policy management from the central console and reporting and graphs
(Crystal Reports).
RealSecure 7.0 was released after this article was written.
Dragon 5.0.2
Dragon supports Linux, Solaris Sparc, Solaris x86, HP UX, FreeBSD,
and OpenBSD. This Unix-based product and the console requires (similar
to Snort) a basic Web server. All alerts and reporting are displayed within
a Web browser (Figure 3). Dragon provides a low-level interface for editing
and configuring signatures; a feature for automatic signature updates
is available. Commercial support is included in the base contract.
|
Figure 3. Enterasys Networks’ Dragon has a Web
interface for creating verbose and graph reports on the captured data.
(Click image to view larger version.) |
Most of the components (including the database) shipped with Dragon are
proprietary. The exception is a Web server, which is required for Dragon
installation. All communications between the agents and console are encrypted.
Users must provide a license key during the installation.
Most of the signatures for the known attacks are proprietary and must
be downloaded from the vendor via a remote update. Users can push policies
(containing signatures and rules) to remote NIDS sensors, thus allowing
for centralized remote administration. Among its features are automated
manipulation of network traffic (connection resets).
Snort 1.8.7b121 with Analysis
Console for Intrusion Detection 0.96b21
Snort is an open source tool that supports Windows 95, 98, Me and
XP; NT; Win2K; Solaris; FreeBSD; OpenBSD; and Linux. Source code is also
provided for compilation on other platforms.
Snort is a free IDS partially based on the TCPDump. On its own it can
sniff traffic, log traffic to a text file or log alerts. It can also log
to a database, create SNMP traps, log to an XML file, post to an HTTP
server, send SMB alerts, log to the EventLog and log to a syslog server.
It can be combined with the Analysis Console for Intrusion Detection (ACID)
console (Figure 4), MySQL, PHP, and a Web server to add more robust alert
reporting and analysis. The open source community has developed a number
of tools to complement and extend Snort’s native capabilities. Signatures
for publicly disclosed vulnerabilities are available from the Snort Web
site and third-party sites such as the www.whitehats.com “arachNIDs” database.
The Snort rules scripting language facilitates the creation of new signatures
and the customization of existing signatures.
|
Figure 4. Working in concert with Snort, ACID
can conveniently group attacks by classification, reporting sensor
and source and destination IP address. (Click image to view larger
version.) |
If your company prefers commercial support, it’s available from Silicon
Defense, www.silicondefense.com. Also, Source Fire, www.sourcefire.com,
provides support for Snort-based hardware appliances. Free support is
available via Internet mailing lists.
Installation of Snort requires a Web server supporting Perl and PHP,
the ACID console (a number of PHP scripts), database for alert logging,
and WinPcap packet capture driver. Comprehensive Windows installation
instructions are available from multiple sources listed at www.silicondefense.com/techsupport/downloads.htm.
The Tale of the Tape
Our product tests used evaluation versions of all software, except
Snort. Dragon 5.0.2 provides a limited number of signatures and capabilities
in its evaluation version. All tests were performed using additional sniffers
and session loggers to ensure that expected traffic was, indeed, passed
between the attacker and Web server.
All NIDS detected TCP stealth scans performed by nmap. Snort identified
nmap as the tool used for the port scan, while other products merely established
the fact that port scans were performed. Dragon didn’t report any alarms
associated with the UDP scan, while RealSecure, NetProwler and Snort triggered
several alerts related to UDP scanning, unsolicited DNS replies (NetProwler)
and UDP-based attacks (Snort). UDP scan alert on the three out of four
products was a correct alarm. The rest of the reported events can be considered
false positives in this case (or extra “noise”), as no actual UDP attacks
were executed as part of this activity. Only Snort accurately detected
a ping sweep and established that it was performed by nmap. While port
scans are common on the Internet, an activity such as this may contradict
corporate policy and require an investigation if detected on the internal
network.
All four products easily identified a vulnerability scan executed by
ISS Network Scanner 6.2.1. However, only RealSecure and Dragon identified
ISS Network Scanner as the agent probing for vulnerabilities and attacks
against the target system. Snort reported more than 100 alerts in the
first two minutes of the scan, and NetProwler identified a port scan as
well as multiple TFTP and FTP attacks. The fact that some of the products
are able to accurately recognize the signatures of the tools used to probe
for vulnerabilities or execute the attacks may prove useful in the validation
of events and further escalation and investigation.
All four products detected a massive 30-minute Nessus scan via the Internet.
RealSecure and Dragon logged 21 and 13 unique alert signatures, respectively,
precisely identifying some of the attacks (such as SNMP vulnerabilities
specific to HP OpenView and Cisco’s IOS). NetProwler caught just a handful
of attacks related to Back Orifice (a popular Trojan back door), TFTP,
FTP, as well as port scans. Snort generated more than 2,000 alerts related
to about 30 unique signatures.
RealSecure, NetProwler and Snort identified probes for common gateway
interface (CGI) vulnerabilities generated by Arirang. The evaluation version
of Dragon remained silent. Since port 80 (HTTP) remains open on the majority
of corporate firewalls, it’s critical to ensure that CGI scripts on the
Web servers have strong security. You’ll also want to know if probing
for specific CGI components applicable to your environment takes place.
The test using Stick proved the point that IDS is subject to obfuscation
through flooding. All four tools generated thousands of alerts within
seconds. However, because Stick is designed to send the IDS rule base
toward the target, it comes as no surprise and proves only one point:
Depending on the flow of traffic visible to the NIDS, as well as traffic
filtering patterns around the NIDS, alert reporting and processing may
become unmanageable. Note, however, that IDSs are being improved to become
much more resistant to Stick.
One other useful application of Stick is the ability to stress-test your
IDS system. Because Stick is designed to produce as much traffic as the
hardware of the source system will allow, it’s possible to establish the
CPU utilization, packet drop rate and other important limits of the NIDS.
The “fragroute” test showed that RealSecure can reassemble fragmented
attack packets correctly and still report on the same attack signatures
generated by Stick. Dragon didn’t pick up any attacks, but did generate
hundreds of informational messages reporting on the “odd TCP flag combinations.”
Therefore, instead of correctly reassembling the packets, Dragon identified
“protocol anomaly” in the network traffic and alluded to suspicious activity.
In a production environment, protocol anomaly validation and investigation
can be incredibly difficult and time consuming.
Our tests determined that all the products are effective in detecting
and reporting on suspicious traffic and malicious attacks. In choosing
your IDS solution, you need a clear understanding of product features
and functionality, along with your organization’s objectives of the implementation
(simple detection of known attacks vs. network traffic monitoring for
operational and troubleshooting purposes). An organization should assess
whether it’s able to manage and respond to certain events internally or
whether it should rely on a managed service provider. Finally, the IT
department may choose to test and evaluate the product on its own or rely
on technical evaluations provided by various research labs (such as Neohapsis
at www.neohapsis.com/neolabs/default.php
or Miercom at www.mier.com/
testserv.html).
And the Winner is…
In our opinion, Snort stands out in the evaluation. It’s a system
embraced by many security professionals, and it comes with powerful interface
additions and a large number of signatures that can be tailored to your
requirements. Support for Snort is also available through commercial providers
like Source Fire.
ISS, publisher of RealSecure, appears to have improved in its effectiveness
and product support over the last year. In our experience, ISS products
are easy to use and equipped with useful features, though by means of
proprietary technology and in some cases at the expense of performance.
NetProwler appears to be the most outdated NIDS product. It’s especially
alarming that its critical components are limited to NT 4.0, as Microsoft
is phasing out sales and support of the OS. If Symantec is serious about
NetProwler, it should be at work on a next-generation release. Dragon
proved to be a solid system featuring Java-based Web reporting interfaces.
Enterasys just recently released version 6 of the software. Dragon goes
beyond the NIDS arena and today provides a full suite of IDS features,
including third-party IDS managed services.
12
Steps to Successful IDS Deployment |
- Define your objectives and establish a preliminary
budget. Identify critical systems and/or network segments
that you need to monitor. Establish what you’re looking
for and where such traffic can be captured in its
entirety. Based on this analysis, you’ll be able to
define the scope of your implementation as well as
start the product evaluation process. In many cases,
budget will drive further product selection.
- Plan your IDS deployment. Remember that a successful
NIDS deployment should be part of the overall security
strategy. Your IT department has to budget for both
the cost of implementation as well as tune-up, monitoring,
alerting and response procedures. Decide whether to
build this expertise and process in-house, outsource
it or do both.
- Select the products based on what’s important to
your organization. You should at least consider: a)
placement preferences and performance requirements
(line speeds); b) business applications to monitor
(Web, e-mail or general); c) threats and technologies
to detect (protocols, internal/external access); d)
degree of signature customization available; e) reporting
mechanisms; and f) available real-time alert notification
methods.
- Architect your NIDS deployment. Decide whether to
deploy a management network or integrate with the
main network. Design encryption channels and access
controls for reporting of alerts, build redundancy
in the reporting infrastructure, and establish channels
for console reporting as well as automated pager/e-mail
alert notification.
- Prioritize the alerts you want your NIDS system
to generate.
- Establish incident response procedures. Without
incident response, NIDS deployment may turn out to
be useless or just be used to collect forensics data.
All parties involved in the NIDS management and monitoring
have to have a very clear understanding of what action
to take based on a specific type of alert. The incident
response procedure should specify actions to take
in case of NIDS obfuscation events, as well as in
the case of critical security events. The procedures
should clearly state when or if to call law enforcement
and the possibility of discontinuing IT operations
in case of a serious compromise. Finally, the procedures
should state requirements for data collection and
retention for forensic analysis, if needed. You can
obtain more information on incident response procedures
from CERT at www.cert.org/nav/index_red.html.
- Conduct a pilot implementation. At this stage,
you should still be able to evaluate some of the products
and test your hypothesis regarding security events
you’re trying to detect. You can also test your incident
response procedures at this point.
- Train personnel responsible for NIDS deployment,
management and administration. If the NIDS is deployed
in-house, provide extensive training at all levels.
If the NIDS is outsourced, associates serving as liaisons
with the vendor should be trained in NIDS technologies
and incident response procedures.
- Deploy NIDS.
- Test your installation. Industry averages show
it takes two to 10 months to tune a NIDS implementation.
Continuous testing, data analysis and configuration
adjustments ensure your NIDS deployment works as expected.
- Proactively monitor news and vendor announcements
for new threats and attacks. Perform timely updates
to your NIDS installation to detect all attacks and
anomalies of interest.
- Assess your threats and evaluate the adequacy of
your NIDS deployment as your network topology continues
to change.
|
|
|
Keeping the Barbarians at Bay
Intrusion detection systems are starting to achieve the level of
ubiquity in the corporate environment that firewalls attained in the late
1990s. More and more, NIDS are seen as an important component of a secure
e-business infrastructure—as proven by the number of regulatory agencies
and industry groups requiring their presence. In the U.S., this is especially
true of the financial services and healthcare industries, as well as other
industries that are part of the “critical infrastructure.” While we’ve
given a broad overview of IDS and sampled some of the better known products,
this is really just an introduction to the topic. If you have security
responsibilities on your network, you owe it to your employer or clients
to delve deeper into intrusion detection and learn all you can about this
growing and increasingly crucial field.
Greg Saoutine, MSCE, CISSP, CCNA, Sun Certified System
Administrator, has spent several years as a security consultant. Currently,
Greg is an IT security manager at a Fortune 500 company.
Sergey Perfiliev, MCSE, CCDP, Sun Certified System Administrator,
is
a Certified Information Systems Auditor with 10-plus years of developing,
administering and auditing client-server and distributed systems. His
interests focus on emerging issues in information security.
Robert Corti, MCSE, CCNA, CCDA, CISSP, has spent several years as
a New
York City-based security consultant for Fortune 1000 companies. Currently,
Robert is a risk management professional for a New York-based technology
company.