2011 IBM X-Force Trend and Risk Report

IBM X-Force® 2010
Trend and Risk Report

March 2011
IBM Security Solutions
IBM X-Force® 2010 Trend and Risk Report
Dedication
Dedication
The IBM X-Force® 2010 Trend and Risk Report is dedicated in memory of our friend and colleague Bryan
Williams who passed away during this effort. His knowledge and focus on the changing threat landscape of
virtualization is documented in this report. Bryan was a highly valued member of the IBM X-Force team
since the early days and his contribution to the team, security and IBM are too numerous to list. He will be
greatly missed.
IBM Confidential and Internal only—2011 2

Contributors
Contributors
Producing the X-Force Trend and Risk Report is a dedication in collaboration across all of IBM. We would About X-Force
like to thank the following individuals for their rapt attention and dedication to the publication of this report. The IBM X-Force® research and development teams
study and monitor the latest threat trends including
Contributor Title
vulnerabilities, exploits and active attacks, viruses
Amrit Williams Director, Emerging Security Technology and other malware, spam, phishing, and malicious
Bryan Williams X-Force Research and Development, Protection Technologies web content. In addition to advising customers and
Carsten Hagemann X-Force Software Engineer, Content Security the general public about emerging and critical
Colin Bell Principle Consultant, AppScan OnDemand Services threats, X-Force also delivers security content to help
David Merrill STSM, IBM Chief Information Security Office, CISA protect IBM customers from these threats.
Dr. Jens Thamm Database Management Content Security
Harold Moss Emerging Tech & Cloud Computing Technical Architect
Jay Radcliffe Senior Threat Analyst, MSS
Jeffrey Palatt Manager, Emergency Response Services
John Kuhn Senior Threat Analyst, MSS
Jon Larimer X-Force Advanced Research, Malware
Leslie Horacek X-Force Threat Response Manager
Lisa Washburn Global Product Mgr, IBM Security Services—Threat/Cloud
Marc Noske Database Administration, Content Security
Mark E. Wallis Senior Information Developer for IBM Security Solutions
Matthew Ward Senior Product Manager—Tivoli Security
Michelle Alvarez Team Lead, MSS Intelligence Center(aka Eagle Eyes)
Mike Warfield Senior Wizard, X-Force
Ory Segal Security Products Architect, AppScan Product Manager
Patrick Vandenberg Manager, Rational Security & Compliance Marketing
Ralf Iffert Manager X-Force Content Security
Ryan McNulty IBM Managed Security Services & SQL Querier Extraordinaire
Scott Moore X-Force Software Developer and X-Force Database Team Lead
Shane Garrett X-Force Advanced Research
Steven Bade STSM Security Architect and Strategist
Tom Cross Manager—X-Force Strategy and Threat Intelligence
Wangui McKelvey X-Force Marketing Manager 3
Contents > Section I
Contents
Section I
Dedication 2 Trending in the dark—what does malicious traffic look like? 24 Phishing 57
Contributors 3 Spoofed Denial of Service attacks 24 Phishing volume 57
About X-Force 3 Targets of Denial of Service attacks 26 Are phishers becoming skimmers? 58
Navigating the report 6 Worms of yesteryear: Where are they now? 27 Phishing—country of origin 59
Section I—Threats 7 Web content trends 31 Phishing—country of origin trends 60
Executive overview 7 Analysis methodology 31 Phishing URLs—country of origin 61
2010 Highlights 8 Percentage of unwanted Internet content 32 Phishing URLs—country of origin trends 62
Threats 8 Malicious websites 37 Phishing—most popular subject lines 63
Operating Secure Infrastructure 8 Spammers focus on content rather than volume 42 Phishing targets 64
Developing Secure Software 9 Major content trends in spam for 2010 42
Emerging Trends in Security 10 Spam volume 45
IBM Security collaboration 10 Conclusions about spam volume and content 45
IBM Managed Security Services—A global threat landscape 11 Spammers on holiday at the end of the year 46
Trojan Bot networks 11 Regional spam volume per day of the week 47
SQL injection 13 Common domains in URL spam 48
Obfuscation 15 Common top-level domains in URL spam 51
PDF exploitation 16 Internationalized country code top-level domains:

First occurrences in spam 51
Cross-site scripting 17
Spam—country of origin 52
Industry trends 18
Spam—country of origin trends 54
Top high-volume signatures—IBM MSS 21
Spam URLs—country of origin trends 55
Targeting SMB Servers 22
SQL injection—high volume 23
PsExec 23
Brute force attacks & scans 23
JScript & UNIX 23
4
Contents > Section II, III and IV
Contents
Section II, III and IV
Section II—Operating Secure Infrastructure 68 Virtualization—risks and recommendations 90 Section III—Developing Secure Software 101
Advanced persistent threat (APT) and targeted attacks 68 Virtualization system components 90 Further analysis on web application trends 101
Background and definitions 68 Vulnerability distribution 92 Conclusions from real-world web application assessments 101
Response and research 68 Attacks unique to virtualization systems 93 Hybrid analysis sheds light on vulnerability blind spot 111
Conclusions and recommendations 70 Public exploits 94 Web application hack-ability and efficient defense 114
Stuxnet and SCADA 72 Summary of security concerns 94 Avoid the Net cast by automation 119
Who is behind Stuxnet? 72 Operating Secure Virtual Infrastructure 94 Fix vulnerabilities efficiently 119
Works cited 74 Endpoint security and systems management 96 The best defense against the elite 119
Public vulnerability disclosures in 2010 74 A well-managed device is a more secure device 96 Section IV—Emerging Trends in Security 120
2010—A record setting year 75 The State of Affairs in DNSSEC 98 Mobile security trends 120
Public exploit disclosure 78 Introduction 98 Effective controls to manage mobile devices 122
Vendor supplied patches 79 2010 The year in review 98 Encryption 123
Toward more reliable public vulnerability reporting 80 Software deployment and components 98 Remote Access Service 124
Shift from local to remotely exploitable vulnerabilities 81 DNSSEC challenges and stumbling blocks 99 Future security vision 125
Web application vulnerabilities 82 What’s ahead now 100 The evolving state of security in the cloud 126
Web application platforms vs. plug-ins 84 Conclusion 100 Design elements for security in the cloud 128
Client-side vulnerabilities and exploits 85
Exploit effort versus potential reward matrix 88
Key Recommendations 89
5
Navigating the report
Navigating the report

Welcome. This year we have made some helpful We start by talking about the threats that our
improvements to the format and content of the systems and networks are facing, because we have
Trend Report. These improvements are aimed at to begin by understanding the problem we are all
enabling readers to take the findings a step further. working to solve. Once a threat is understood, we
We understand that computer and network security can work towards realistic technology controls and
is about focusing on awareness of the threat and educational awareness to help secure our enterprise
helping to protect the systems and networks from and systems. In both the Operating Secure
these threats. But then what? As an organization Infrastructure and Developing Secure Software
matures in its stance on computer security and sections we not only discuss threats but provide
known threats, how can they begin to develop a logical advice on how to help improve or detect
deeper focus towards improvement? those threats in your environment. In the Emerging
Trends in Security section, we take a forward look
We asked ourselves that question and determined into emerging technologies that are pressing into
the answer was to provide to our readers a deeper discussions as future business concerns.
understanding of what we experience and have
learned from the breadth of capabilities that is IBM We believe this new layout better organizes the
Security Solutions. material we want to present and helps you the
reader focus on what is most important to your
For this report we have divided the content into four organization.
sections.
• Threats
• Operating Secure Infrastructure
• Developing Secure Software
• Emerging Trends in Security
6
Section I > Executive overview
Section I–Threats
In this section we explore topics that comprise generally, we have seen a rise in hactivism across In our advanced persistent threat article, we look
“Threats” and describe the attacks aimed at the the globe, where attackers are no longer motivated at some of the most sophisticated adversaries our
enterprise that security specialists face. We address simply by self-recognition or financial gain, but by networks have ever faced. These types of low and
the malicious activity observed across the spectrum political change and protest. slow coordinated attacks are often an indicator of
by IBM and how we go about helping protect highly cohesive and organized groups of attackers
networks from those threats. In addition, an update The second half of 2010 also marked a highly who use a variety of sophisticated attack
on the latest attack trends as identified by IBM. visible precedent in the industrial and manufacturing techniques to inch their way into the enterprise.
space. The multi-faceted and highly customized
Executive overview Stuxnet worm shook up the SCADA world by Not only are attacks changing but so is the very
The second decade of the twenty first century is proving how security vulnerabilities can cripple a technology that we utilize to carry this traffic. We
underway and technology continues to permeate factory or production site. No longer is just take a quick look at how networks are scrambling
every aspect of our work and personal lives. At IBM e-commerce, personal, or corporate data at risk, to keep up with technology changes. At the mid-
we call this the Smarter Planet and we are but the very infrastructure that powers our factories year point, we discussed a shift from IPv4 into IPv6
continuously helping our customers to take and energy sector can be exposed for exploitation. requirements and in this report, we discuss the
advantage of a world that’s more interconnected, oncoming advent of DNSSEC.
intelligent, and instrumented. As much as these On a smaller scale, mobile devices continue to
innovations can increase our efficiency and ability to multiply in the workplace, helping increase the 2010 was a pivotal year on many counts and has
instantly connect on a global scale, so too can the magnitude and complexity of risk in protecting the shown that understanding the trends of the security
risks and dangers of a connected world become enterprise. In the emerging trends in security landscape is more critical than ever. IBM continues
more sophisticated and difficult to contain. section, we look at several mobile vulnerabilities its dedicated effort to educate, inform, and discuss
that may be an indicator of more to come. In the security topics and emerging trends with the
To prove the point, the confluence of this innovation enterprise, and at home, web vulnerabilities community at large. Preparing organizations to not
recently showed its face in several authoritarian targeting the browser continue to dominate the only understand the emerging threat landscape, but
countries, where technology and political activism majority of weaknesses, demonstrating the also to better understand the weaknesses of an
have united to empower people in sharing a voice importance of patch compliance and host organization’s infrastructure.
and making change on a global scale. More protection. We discuss an interesting case study of
how large complex organizations can benefit from
centralized patch management.
7
Section I > 2010 Highlights > Threats > Operating Secure Infrastructure
2010 Highlights • Obfuscation, whereby attackers attempt to hide • There were a few months with ups and downs in
Threats their activities and disguise their programming, the volume of spam seen over the year, however,
Malware and the Malicious web continued to increase over 2010 and shows no the overall trends stayed flat and we have seen
• IBM Managed Security Services (MSS) saw an signs of waning. even less volume at the end of the year in
upward trend in Trojan botnet activity during • Compromise through PDF exploitation continues comparison to the beginning of 2010.
2010. This growth is significant because despite to be a favorite among attackers. In late April, a • At 15.5 percent, India was the top country for
increasing coordinated efforts to shut down botnet particular spam campaign contained an Adobe phishing email origination in 2010, followed by
activity (as seen with the Mariposa, Bredolab and Acrobat PDF that used the Launch command to Russia at 10.4 percent.
Waledec botnets), this threat appears to be deliver malware. At the peak of the attacks, IBM • In 2010, financial institutions continue to climb as
gaining momentum. Managed Security Services (MSS) received more the number one target for phishing attempts,
• IBM’s data illustrates the dramatic impact of a than 85,000 alerts in a single day. representing 50 percent of the targeted industries
successful effort in early 2010 to shutdown the • The SQL Slammer worm first surfaced in January up from the mid-year report when it was 49 percent.
Waledac botnet, which resulted in an instantaneous 2003 and became known as one of the most • In 2010, more than three out of four financial phishing
drop off in observed command and control traffic. devastating Internet threats of the past decade. emails targeted banks located in North America.
• Zeus (also known as Zbot and Kneber), continues This worm continued to generate a great deal of The remaining 22 percent targeted Europe.
to evolve through intrinsic and plugin advances. traffic on the Internet in 2010.
The Zeus/Zbot family of botnets has been around Operating Secure Infrastructure
for many years now and due to its extreme Web content, spam, and phishing Vulnerabilities and Exploitation
popularity with attackers, there are hundreds, or • IBM Content security team identified that in the • According to the X-Force database tracking, 2010
even thousands, of separate Zeus botnets active past three years, anonymous proxies have steadily had the largest number of vulnerability disclosures
at any given time. The Zeus botnet malware is increased, more than quintupling in number. in history—8,562. This is a 27 percent increase
commonly used by attackers to steal banking Anonymous proxies are a critical type of website over 2009, and this increase has had a significant
information from infected computers. to track, because they allow people to hide operational impact for anyone managing large IT
• SQL injection is one of the leading attack vectors potentially malicious intent. infrastructures. More vulnerability disclosures can
because of its simplicity to execute and its • USA, India, Brazil, Vietnam, and Russia are the top mean more time patching and remediating
scalability to compromise large amounts of web five countries for spam origination in 2010. vulnerable systems.
servers across the Internet. There also appears to • In 2010, spammers focused on content over
be a seasonal pattern: during each of the past volume. At the beginning of August, spammers
three years, there has been a globally scaled SQL began sending spam threats with ZIP attachments
injection attack some time during the months of that contained a single EXE file that was malicious.
May through August. By September, spammers began shifting to HTML
spam to once again trick the end-user.
8
Section I > 2010 Highlights > Developing Secure Software
• 49 percent of the vulnerabilities disclosed in 2010 Virtualization • ASP.NET applications were clearly more
were web application vulnerabilities. The majority of • IBM X-Force notes that virtualization systems susceptible to SQL injection than Java or PHP.
these were cross site scripting and SQL injection added 373 new vulnerabilities to the network The likely reason is that applications would typically
issues. However, as IBM X-Force has been saying infrastructure in the period between 1999 use SQL Server as a backend database. SQL
for years, these vulnerabilities represent just the tip and 2009. injection is better documented and easier to detect
of the iceberg since many organizations develop • A number of public exploits exist that in this technology.
third-party applications in-house that are never demonstrate the risk from virtualization • As Web 2.0, AJAX applications, and Rich Internet
even reported publically and are not included in system vulnerabilities is real. Applications (RIAs) become more common,
this count. • Hypervisor escape vulnerabilities are the most client-side JavaScript vulnerabilities may become
• Although vendors have been diligent in providing common type of vulnerability that has been more relevant, with a potential rise in the amount of
patches, at least 44 percent of all vulnerabilities in disclosed in server class virtualization systems. such issues being exploited by malicious attackers.
2010 still had no corresponding patch by the end • A recent IBM research study discovered that about
of the year. Developing Secure Software 14 percent of the Fortune 500 sites suffer from
• In early 2010, the term Advanced Persistent Threat Web Application Vulnerabilities many severe client-side JavaScript issues, which
(APT) became part of the everyday information • From the IBM® Rational® AppScan® OnDemand could allow malicious attackers to perform attacks
security lexicon as a result of certain public Premium Service we observed web application such as
disclosures and acknowledgement of a targeted vulnerabilities comprising 49 percent of the total –– Infecting users of these sites with malware
series of attacks known as Operation Aurora. vulnerabilities reported in 2010, it is no surprise and viruses.
There has been much debate over this term and that developing secure software is harder –– Hijacking users’ web sessions and performing
the underlying concepts within the information than ever. actions on their behalf.
security community. • In 2010 for the first time we now find that Cross- –– Performing phishing attacks on users of
• During certain public disclosures in early 2010, and Site Request Forgery (CSRF) is more likely to be these sites.
after attacks associated with Operation Aurora, the found in our testing than Cross-Site Scripting –– Spoofing web contents.
term APT began to take on a different meaning. In (XSS). This change is attributed to better detection
• Based on the dataset that we analyzed, we may
essence, APT became associated with any techniques for CSRF and also a greater awareness
extrapolate that the likelihood that a random page
targeted, sophisticated, or complex attack of the risk. We find that organizations will tolerate
on the Internet contains a client-side JavaScript
regardless of the attacker, motive, origin, or having some outstanding issues with CSRF if the
vulnerability is approximately one in 55.
method of operation. risk of exploitation is minimized. This is not the
case with XSS and these issues are often
quickly resolved.
9
Section I > 2010 Highlights > Emerging Trends in Security > IBM Security collaboration
Emerging Trends in Security

IBM Security collaboration
Mobile
• Mobile devices represent opportunities for IBM Security represents several brands that provide a broad spectrum of security competency.
sophisticated, targeted attackers. There are a • IBM X-Force® research and development teams discover, analyze, monitor, and record a broad range of
number of vulnerabilities to target, and there is computer security threats and vulnerabilities
exploit information available.
• However, it is important to keep the vulnerability • IBM Managed Security Services (MSS) is responsible for monitoring exploits related to endpoints, servers
increases in perspective -- these do represent (including web servers), and general network infrastructure. MSS tracks exploits delivered over the web as
shared software components used by both mobile well as other vectors such as email and instant messaging.
and desktop software. The vulnerability research
• Professional Security Services (PSS) delivers comprehensive, enterprise-wide security assessment, design,
that is driving these disclosures is not necessarily and deployment services to help build effective information security solutions.
mobile-centric.
• Still, we aren’t seeing widespread attack activity • Our Content security team independently scours and categorizes the web through crawling, independent
targeting mobile vulnerabilities today, because discoveries, and through the feeds provided by MSS. In addition, the team actively monitors millions of email
mobile devices do not represent the same kind of addresses to receive mass amounts of spam and phishing emails. This work provides optimal spam
financial opportunity that desktop machines do for protection accompanied by the latest trends in spam and phishing emails.
the sort of individuals who appear to create large •
IBM has collated real-world vulnerability data from security tests conducted over the past three years from the
Internet botnets. IBM® Rational® AppScan® OnDemand Premium Service. This service combines application security
assessment results obtained from IBM Rational AppScan with manual security testing and verification.
Cloud security
• While security is still considered one of the major • IBM Cloud Security Services allows clients to consume security software features through a hosted
inhibitors to cloud adoption, organizations are subscription model that helps reduce costs, improve service delivery, and improve security.
increasingly adopting cloud-based technologies to
• Identity and access management solutions provide identity management, access management, and user
address competitive market needs.
compliance auditing. These solutions centralize and automate the management of users, authentication,
• Extending existing security policies and standards,
access, audit policy, and the provisioning of user services.
leveraging sound physical security protections
already in place, and assessing systems and • IBM Endpoint Management Solutions combine endpoint and security management into a single offering that
applications for security weaknesses are examples enables customers to see and manage physical and virtual endpoints—servers, desktops, roaming laptops,
of security design elements that should be included and specialized equipment such as point-of-sale devices, ATMs and self-service kiosks.
when establishing a secure cloud environment.
10
Section I > IBM Managed Security Services—A global threat landscape > Trojan Bot networks
IBM Managed Security Services— availability of bot exploit toolkits such as WARBOT. Trojan Bot networks also continued to evolve in
A global threat landscape This allows less than tech-savvy individuals to take 2010. One of them, Zeus (also known as Zbot and
IBM Managed Security Services (MSS) monitors advantage of the lucrative business of selling Kneber), continues to evolve through intrinsic and
several billion events in more than 130 countries, 24 sensitive information on the black market. plugin advances. The Zeus/Zbot family of botnets
hours a day, 365 days a year. The global presence of
IBM MSS provides a first-hand view of current threats. Trojan Botnet Activity
IBM analysts use this wealth of data to deliver a 250,000
unique understanding of the cyber threat landscape.
This section focuses on Trojan botnet activity, SQL
injection, obfuscation, PDF exploitation, and 200,000
cross-site scripting activity—threats that are
discussed throughout this report. The trend of
these threats is vital to determining what direction 150,000
Event Count
the threat is taking and to understanding the
significance of the threat to our networks.
100,000
Trojan Bot networks
IBM MSS saw an upward trend in Trojan botnet
activity during 2010. This growth is significant 50,000
because despite increasing coordinated efforts to
shut down botnet activity (as seen with the
Mariposa1 and Bredolab2 botnets), this threat 0
1/1/2010
1/12/2010
1/23/2010
2/3/2010
2/14/2010
2/25/2010
3/8/2010
3/19/2010
3/30/2010
4/13/2010
4/24/2010
5/5/2010
5/16/2010
5/27/2010
6/7/2010
6/18/2010
6/29/2010
7/10/2010
7/21/2010
8/1/2010
8/12/2010
8/23/2010
9/3/2010
9/14/2010
9/25/2010
10/6/2010
10/17/2010
10/28/2010
11/8/2010
11/19/2010
11/30/2010
12/11/2010
12/22/2010
appears to be gaining momentum. While there have
been some successful shutdowns there are many
botnets that, due to their resilient and sophisticated
Command and Control (CnC) topology, remain
Total Trend Line
largely unaffected by these takedown attempts.
Another reason attributing to this growth is the
Figure 1: Trojan Botnet Activity
1 Massive Mariposa botnet shut down – http://www.net-security.org/secworld.php?id=8962

2 Bredolab botnet shut down – http://nakedsecurity.sophos.com/2010/10/26/bredolab-botnet-shut/ 11
Section I > IBM Managed Security Services—A global threat landscape > Trojan Bot networks
has been around for many years now and due to its the application, but certain versions of Foxit Reader dramatic drop? We speculate that the cessation in
extreme popularity with attackers, there are do not and merely start the application without user activity is the result of “Operation b49”.5 This Microsoft
hundreds, or even thousands, of separate Zeus confirmation. In cases where organizations have led operation resulted in the takedown of a majority
botnets active at any given time. The Zeus botnet moved away from Adobe’s implementation, this is of this botnet in late February. Once a temporary
malware is commonly used by attackers to steal of particular concern with regards to this attack. restraining order was granted on February 22nd,
banking information from infected computers. much of the communication between Waledac’s
Zeus’ encrypted command and control activity is hard command and control centers and its thousands of
Various bot networks based on Zeus are to detect. However, one of the signatures analyzed to zombie computers was cut off in a matter of days.
responsible for millions of dollars in losses over the assess this threat focuses on a type of behavior that In October, the U.S. District Court of Eastern
last few years. For example, Zeus was reportedly Zeus might exhibit. The signature HTTP_Suspicious_ Virginia ordered the permanent transfer of
responsible for stealing more than $1 million from Unknown_Content detects when a HTTP POST ownership of the 276 domains behind Waledac to
customers of a single UK-based financial institution message results in a session where the content sent Microsoft.6 Does this mean that Waledac will never
in July.3 The continual arms race between attackers and received is not recognized as typical content, surface again? We may see activity, but probably
and defenders has botnet controllers finding such as images or documents. Activity associated not to the same magnitude that we observed prior
stealthier ways to keep their bots under the radar. with this signature seemed to grow in intensity towards to the takedown.
Zeus’ merger with SpyEye, a very similar Trojan, is the latter half of 2010. Such activity could be normal or
still in its infant stages. How this plays out over time could indicate botnet activity. While this is a generic Another prevalent botnet is Pushdo (also known as
is to be determined, but consolidation amongst signature, we do believe that this activity is associated Pandex and some components are known as
Trojan botnets is expected to be an emerging trend. with Zeus. The section titled “Zeus botnet—facts, Cutwail). This botnet generated noticeable activity
myths and understanding how these botnets operate” across the IBM MSS network in 2010 though to a
In April, we saw a spike in malicious PDF activity in the 2010 Mid-Year Trend and Risk Report lesser extent than Waledac and Zeus. Pushdo, primarily
associated with Zeus.4 Attackers abused the provides an in-depth explanation of Zeus and how used for spamming, had been observed launching
“Launch” feature in Adobe Acrobat to distribute the readers can protect themselves from this threat. Distributed Denial of Service (DDoS) attacks against
Zeus botnet malware via email. The signature certain SSL-enabled websites beginning in the first
PDF_Launch_Program detects the network transfer There was also significant activity associated with the quarter 2010. The DDoS attack involved sending
of a PDF file containing an embedded action to Waledac botnet at the start of the year up until early thousands of malformed SSL requests to the target
Launch an executable program. Adobe Reader March and then the activity seemingly disappears hosts in an attempt to use up resources. To a business,
asks for user confirmation before actually launching for the rest of 2010. What could have caused this this could directly impact revenue if services provided
or product sales are interrupted during such an attack.
3 Targeted Attack Nets 3,000 Online Banking Customers – http://www.darkreading.com/smb-security/security/attacks/showArticle.jhtml?articleID=226600381
4 PDF-based Zeus attacks – http://www.iss.net/threats/PDFbasedZeusAttack.html
5 Cracking Down on Botnets – http://blogs.technet.com/b/microsoft_on_the_issues/archive/2010/02/24/cracking-down-on-botnets.aspx
6 R.I.P. Waledac – Undoing the damage of a botnet http://blogs.technet.com/b/microsoft_blog/archive/2010/09/08/r-i-p-waledac-undoing-the-damage-of-a-botnet.aspx 12
Section I > IBM Managed Security Services—A global threat landscape > SQL injection
SQL injection SQL Injection Attacks Monitored by

SQL injection is one of the leading attack vectors IBM Managed Security Services
seen because of its simplicity to execute and its 45,000
scalability to compromise large amounts of web
servers across the Internet. A review of past
35,000
X-Force Trend and Risk Reports reveals an
interesting SQL injection trend. During each of the
past three years, there has been a globally scaled 25,000
SQL injection attack some time during the months
of May through August. The anatomy of these 15,000
attacks is generally the same: they target .ASP
pages that are vulnerable to SQL injection. The
5,000
surges that occurred during 2008 and 2009 are
shown in Figure 2.
May June
2008
In 2008, attackers used a SQL CAST statement and
some hex code to obfuscate the true injection string. 800,000
The source of this attack was the Asprox botnet, 700,000
and it was massively successful in compromising
600,000
thousands of websites. In 2009, we observed the
same attack methodology; the only difference was 500,000
in the resulting payload. Asprox was again the 400,000
source of this attack. However, it had varied 300,000
success this time because of countermeasures that 200,000
were deployed to thwart the attack.
100,000
0
SEP OCT NOV DEC JAN FEB MAR APR MAY JUN
2008 2009
Figure 2: SQL Injection Attacks Monitored by IBM Managed Security Services
13
Section I > IBM Managed Security Services—A global threat landscape > SQL injection
Figure 3 illustrates the significant SQL injection attack SQL_Injection_Declare_Exec Activity

observed in 2010 as detected by the IBM signature
SQL_Injection_Declare_Exec. The same attack 8,000
methodology is used as in the previous two years,
but some of the mechanics were changed. Attackers 7,000
added leetspeak (1337) to the SQL statement to
evade poorly written regex filtering. This statement, 6,000
once decoded, contains another CAST statement
resulting in two layers of obfuscation. While very 5,000
Event Count
similar to Asprox, this attack used slightly different
techniques and therefore is known more popularly 4,000
as the “dnf666” attack—so named because of a
URL encoded inside. 3,000
2,000
1,000
7/14/2010
7/16/2010
7/18/2010
7/20/2010
7/22/2010
7/24/2010
7/26/2010
7/28/2010
7/30/2010
8/01/2010
8/03/2010
8/05/2010
8/07/2010
8/09/2010
8/11/2010
8/13/2010
8/15/2010
8/17/2010
8/19/2010
8/21/2010
8/23/2010
Figure 3: SQL_Injection_Declare_Exec Activity
14
Section I > IBM Managed Security Services—A global threat landscape > Obfuscation
Obfuscation Obfuscation Activity

IBM MSS continues to track trends in obfuscation 400,000
techniques used by attackers and toolkits.
Obfuscation is a technique to hide or mask the 350,000
code used to develop applications. New 300,000
obfuscation methods are constantly evolving in an
Event Count
attempt to evade intrusion prevention systems (IPS) 250,000
and anti-virus which often can’t decode the web
200,000
page or file to find the hidden attack. Through
special detection algorithms incorporated into IBM 150,000
Security Network IPS, we watch how patterns of
100,000
use change by monitoring hits on these algorithms
in our world-wide MSS deployments. 50,000
Obfuscation activity continued to increase during 0

2010 and shows no signs of waning. The most
1/1/2010
1/15/2010
1/29/2010
2/12/2010
2/26/2010
3/12/2010
3/26/2010
4/10/2010
4/24/2010
5/8/2010
5/22/2010
6/5/2010
6/19/2010
7/3/2010
7/17/2010
7/31/2010
8/14/2010
8/28/2010
9/11/2010
9/25/2010
10/9/2010
10/23/2010
11/6/2010
11/20/2010
12/4/2010
12/18/2010
observed activity came from an event that triggers
when a JavaScript ‘unescape()’ function with a
large amount of escaped data is detected. This
activity should be viewed with suspicion. It may be
normal activity, or it could indicate the attempt to Total
inject a large amount of shell code or malicious
HTML and/or JavaScript for the purpose of taking
Figure 4: Obfuscation Activity
control of a system through a browser vulnerability.
15
Section I > IBM Managed Security Services—A global threat landscape > PDF exploitation
PDF exploitation PDF Activity

Compromise through PDF exploitation continues to 200,000
be a favorite among attackers. Throughout 2010, 180,000
160,000
our global Security Operation Centers witnessed
140,000
Event Count
surges of malicious traffic surrounding spam email.
120,000
One notable increase occurred in late April, as 100,000
shown in Figure 5. The emails of this particular 80,000
spam campaign contained an Adobe Acrobat PDF 60,000
that used the Launch command to deliver malware. 40,000
At the peak of the attacks, IBM MSS received more 20,000
0
than 85,000 alerts in a single day. The spam email
7/31/2010
8/5/2010
7/11/2010
7/16/2010
7/21/2010
7/26/2010
5/7/2010
6/6/2010
6/11/2010
6/16/2010
6/21/2010
6/26/2010
7/1/2010
7/6/2010
4/27/2010
5/2/2010
5/12/2010
5/17/2010
5/22/2010
5/27/2010
6/1/2010
3/22/2010
3/27/2010
4/2/2010
4/7/2010
4/12/2010
4/17/2010
4/22/2010
11/3/2010
11/8/2010
10/14/2010
10/19/2010
10/24/2010
10/29/2010
11/13/2010
11/18/2010
11/23/2010
11/28/2010
12/3/2010
12/8/2010
12/13/2010
12/18/2010
12/23/2010
12/28/2010
1/31/2010
2/10/2010
2/15/2010
2/20/2010
2/25/2010
3/2/2010
3/7/2010
3/12/2010
3/17/2010
9/24/2010
9/29/2010
10/4/2010
10/9/2010
1/1/2010
1/6/2010
1/11/2010
1/16/2010
1/21/2010
1/26/2010
2/5/2010
8/10/2010
8/15/2010
8/20/2010
8/25/2010
8/30/2010
9/4/2010
9/9/2010
9/14/2010
9/19/2010
was sent from various SMTP servers globally, which
appeared to originate from the Zeus botnet.
There has been a small but steady rise in PDF Total Trend Line
exploitation since the beginning of 2010. There are
numerous signatures that contribute to this
Figure 5: PDF Activity
assessment. Some of these signatures detect an
unauthorized access attempt. For example, one
signature detects a file with embedded corrupt
JBIG2 data that could cause a buffer overflow in
vulnerable versions of Adobe Acrobat and Adobe
Reader. (Note: This is fixed in Adobe Acrobat/
Reader 8.1.3.) Other signatures may simply be
looking for suspicious activity such as a PDF file
containing a hex-encoded form of a filter name.
This suggests malicious intent by concealing
compressed content within the document.
16
Section I > IBM Managed Security Services—A global threat landscape > Cross-site scripting
Cross-site scripting Cross-Site Scripting Activity

While cross-site scripting vulnerabilities continue to 900,000
be one of the predominant types of vulnerabilities
800,000
affecting web applications, activity targeting these
vulnerabilities seems to have leveled off in 2010 as 700,000
shown in Figure 6. Cross-site scripting allows 600,000
Event Count
attackers to embed their own script into a page the
500,000
user is visiting, thereby manipulating the behavior or
appearance of the page. These page changes can 400,000
be used to steal sensitive information, manipulate 300,000
the web application in a malicious way, or embed 200,000
additional content on the page that can exploit
other vulnerabilities. 100,000
0
Though the trend is flat, it does not mean that this
1/1/2010
1/26/2010
2/20/2010
3/17/2010
4/12/2010
5/7/2010
6/1/2010
6/26/2010
7/21/2010
8/15/2010
9/9/2010
10/4/2010
10/29/2010
11/23/2010
12/18/2010
threat is non-existent. From a Common Vulnerability
Scoring System (CVSS) scoring perspective, these
vulnerabilities do not typically rank as high or critical
threats. IT and security professionals tend to deploy
counter measures for the high-profile vulnerabilities
first and, if resources allow, later address the low- Total Trend Line
to medium-rated issues. Attackers, therefore, will
continue to take advantage of this window of Figure 6: Cross-Site Scripting Activity
opportunity in years to come.
17
Section I > IBM Managed Security Services—A global threat landscape > Industry trends
Industry trends Cross-Site Scripting - Industries with Downward Trend

There is great interest in the general security 80,000
community in knowing which industries are being
70,000
targeted by what attack types. Our customer base
is broad and reaches into a number of different 60,000
industries. However, to identify a valid trend across
a particular industry, we needed to establish a 50,000
methodology with an acceptable sample size for 40,000
analysis. For each attack category, we only
assessed activity where a specific criterion was met 30,000
in a given industry. A minimum number of affected
20,000
customers and a minimum number of devices
deployed amongst those customers was required 10,000
prior to making an assessment.
0
1/1/2010
1/10/2010
1/19/2010
1/28/2010
2/6/2010
2/15/2010
2/24/2010
3/5/2010
3/14/2010
3/23/2010
4/2/2010
4/11/2010
4/20/2010
4/29/2010
5/8/2010
5/17/2010
5/26/2010
6/4/2010
6/13/2010
6/22/2010
7/1/2010
7/10/2010
7/19/2010
7/28/2010
8/6/2010
8/15/2010
8/24/2010
9/2/2010
9/11/2010
9/20/2010
9/29/2010
10/8/2010
10/17/2010
10/26/2010
11/4/2010
11/13/2010
11/22/2010
12/1/2010
12/10/2010
12/19/2010
12/28/2010
Trend Line (Professional, Scientific and Technical Activities)
Figure 7: Cross-Site Scripting – Industries with Downward Trend
18
What did we see? Generally speaking, we did not Cross-Site Scripting - Industries with Downward Trend
see any significant discrepancies across different
industries regarding the varying attack types 80,000
compared to overall customer trends. Attack trends 70,000
across all industries were relatively uniform.
60,000
What can be deduced from this? While some
attacks are targeted, many exploits in circulation 50,000
simply don’t discriminate. A financial organization 40,000
may be just as vulnerable to the latest botnet or
PDF exploitation as an educational institution. 30,000
Whether or not an organization is vulnerable to
attack has much more to do with the protection 20,000
measures that they have in place. 10,000
10/17/2010
10/26/2010
11/13/2010
11/22/2010
12/10/2010
12/19/2010
12/28/2010
1/10/2010
1/19/2010
1/28/2010
2/15/2010
2/24/2010
3/14/2010
3/23/2010
4/11/2010
4/20/2010
4/29/2010
5/17/2010
5/26/2010
6/13/2010
6/22/2010
7/10/2010
7/19/2010
7/28/2010
8/15/2010
8/24/2010
9/11/2010
9/20/2010
9/29/2010
10/8/2010
11/4/2010
12/1/2010
1/1/2010
2/6/2010
3/5/2010
4/2/2010
5/8/2010
6/4/2010
7/1/2010
8/6/2010
9/2/2010
Transportation and Storage Trend Line (Transportation and Storage)
19
The only exception to our findings of consistent Cross-Site Scripting - Industries with Downward Trend
trends among the industries was in the cross-site 120,000
scripting category. As shown in Figure 6, the overall
trend for cross-site scripting was relatively flat and 100,000
several industries followed this trend. As shown in
Figures 7 through 9, a few industries saw a slight 80,000
downward trend in this attack category including:
60,000
• “Professional and Scientific”
• “Wholesale and Retail Trade” 40,000
• “Transportation and Storage”
20,000
A decrease in cross-site scripting activity may
indicate greater attention to addressing these types 0
of vulnerabilities. As noted later in this report, the
1/1/2010
1/9/2010
1/17/2010
1/25/2010
2/2/2010
2/10/2010
2/18/2010
2/26/2010
3/6/2010
3/14/2010
3/22/2010
3/30/2010
4/8/2010
4/16/2010
4/24/2010
5/2/2010
5/10/2010
5/18/2010
5/26/2010
6/3/2010
6/11/2010
6/19/2010
6/27/2010
7/5/2010
7/13/2010
7/21/2010
7/29/2010
8/6/2010
8/14/2010
8/22/2010
8/30/2010
9/7/2010
9/15/2010
9/23/2010
10/1/2010
10/9/2010
10/17/2010
10/25/2010
11/2/2010
11/10/2010
11/18/2010
11/26/2010
12/4/2010
12/12/2010
12/20/2010
12/28/2010
IBM Rational AppScan on Demand Premium
service that tracks web application
vulnerabilities has also seen a steady decline in
the instances of cross-site scripting reported Wholesale and Retail Trade
vulnerabilities since 2007. Part of this decline is
Trend Line (Wholesale and Retail Trade)
attributed to a greater awareness of the risk
associated with cross-site scripting.
20
Section I > Top high-volume signatures—IBM MSS
Top high-volume signatures— Rank Event Name Trend Line

IBM MSS
Table 1 to the right, shows the placement of the top 1 SQL_SSRP_Slammer_Worm Down
MSS high volume signatures and their trend line
for 2010. 2 SQL_injection Down
3 PsExec_Service_Accessed Slightly Up
The top high volume signatures seen across the
MSS network reveal some interesting aspects of life 4 SSH_Brute_Force Slightly Down
on the Internet today and are a reflection of the
longevity of certain threats. For example, the SQL 5 JScript_CollectGarbage Up
Slammer worm7 first surfaced in January 2003 and
became known as one of the most devastating 6 HTTP_Unix_Passwords Slightly Up
Internet threats of the past decade. Despite the
7 SMB_Mass_Login Down
downward trend in 2010, this worm still exists and
continues to propagate as evidenced by the top 8 SMB_Empty_Password No Change
ranking signature, SQL_SSRP_Slammer_Worm
shown in Table 1. SQL Slammer targets a buffer 9 SQL_Empty_Password Up
overflow vulnerability in the Resolution Service in
Microsoft SQL Server 2000 or Microsoft Desktop Table 1: Top MSS high volume signatures and trend line
Engine (MSDE) 2000 installations. This issue was
patched by Microsoft in 2002. The fact that there is
such a huge volume of activity associated with SQL
Slammer seven years after it first surfaced probably
suggests a need for better patch management.
7 SQL slammer traffic on the Internet significantly declined in March 2011 shortly before publication of this report.
For more information on this topic, please see the Frequency-X blog. (http://blogs.iss.net/index.html) 21
Section I > Top high-volume signatures—IBM MSS > Targeting SMB Servers
Targeting SMB Servers as compromised. The SMB_Mass_Login signature connect to SMB servers with no password, this
Two of the top signatures protect against threats detects an excessive number of granted NetBIOS signifies that this method of attack continues to be
targeting server message block (SMB) servers. The sessions originating from the same IP address. This fruitful for attackers. Recent threats, such as the
SMB_Empty_Password detects when a successful may indicate a stolen account being used in a Conficker and Stuxnet malware, use SMB shares to
connection with no password is made to an SMB scripted attack. The existence of these signatures spread across networks.
server. If this connection is from outside the in the list highlights a possible lack of basic security
network, consider the information on your server with SMB shares. If attackers are attempting to
2010 Top 9 High Volume Signatures 2010 Top 8 High Volume Signatures
7E+09 45,000,000
40,000,000
6E+09
35,000,000
5E+09
30,000,000
Event Count
Event Count
4E+09 25,000,000
3E+09 20,000,000
15,000,000
2E+09
10,000,000
1E+09
5,000,000
0 0
Total Total
HTTP_Unix_Passwords SMB_Empty_Password SQL_Injection HTTP_Unix_Passwords SMB_Empty_Password SQL_Injection

JScript_CollectGarbage SMB_Mass_Login SQL_SSRP_Slammer_Worm JScript_CollectGarbage SMB_Mass_Login SSH_Brute_Force
PsExec_Service_Accessed SQL_Empty_Password SSH_Brute_Force PsExec_Service_Accessed SQL_Empty_Password
Figure 10a: 2010 Top 9 High Volume Signatures Figure 10b: 2010 Top 8 High Volume Signatures
22
Section I > Top high-volume signatures—IBM MSS > SQL injection—high volume > PsExec—A remote administration tool > Brute force attacks & scans > JScript & UNIX
SQL injection—high volume Brute force attacks & scans The Darknet data in that mid-year report shows that
Our heuristic SQL signature had the second highest SSH_Brute_Force is another interesting signature in the level of SSH brute force scanning is steadily
volume seen in 2010. This is not surprising because this list. A brute force attack involves an attacker increasing while the MSS data shows that the level of
SQL injection attacks against web applications are trying to gain unauthorized access to a system by brute force attacks against active SSH servers is high.
very common. IBM MSS has observed a seasonal trying a large number of password possibilities. This
surge in SQL injection attacks during the months of signature detects an excessive number of SSH JScript & UNIX
May through August for the past three years as Server Identifications from an SSH server within a JScript_CollectGarbage detects the transfer of a
discussed in the section IBM Managed Security specified timeframe. Through this type of attack, a JScript file containing a call to the function
Services—A global threat landscape. The other malicious individual may be able to view, copy, or CollectGarbage(). CollectGarbage() is part of the
SQL signature noted in Table 1, SQL_Empty_ delete important files on the accessed server or .NET framework but, according to Microsoft, “is not
Password, detects when a successful connection execute malicious code. Organizations can help intended to be used directly from your code.” This
with no password is made to an SQL server. As mitigate brute-force attacks by disabling direct function has been used by attackers and can be
with the SMB_Empty_Password signature, these access to root accounts and using strong indicative of malicious intent. However, it can also
types of connections should be considered usernames and passwords. be used for legitimate purposes.
suspicious if made from outside the network.
We provided an in-depth view on this topic in the Finally, the HTTP_Unix_Passwords signature
PsExec—A remote administration tool 2010 Mid-Year Trend and Risk Report where we detects attempts to access the /etc/passwd file on
The signature in the third spot, PsExec_Service_ explain the nature of a Darknet. A Darknet is a UNIX systems via a web (HTTP) server. While this
Accessed, is notable in that PsExec is a legitimate black-hole network whose addresses are not activity is sometimes authorized, it can sometimes
application. It is a command line based remote allocated to any active legitimate device or service on be suspicious. This is a very old attack, but is still
administration tool. However, worms and advanced the Internet. When an attacker attempts a brute- successful today.
threats also take advantage of PsExec. The “Here force attack on a particular address in the Darknet
you have” worm, for instance, includes a PsExec they never connect to an SSH server because one
tool that allows it to copy itself onto other does not exist. Therefore, they stop after one
computers over the network. If this application is attempt. Conversely, a successful SSH connection
used in your organization, you should ensure that may result in thousands of brute force attempts
best security practices are employed. which explains the large volume of activity
associated with SSH_Brute_Force.
23
Section I > Trending in the dark—what does malicious traffic look like? > Spoofed Denial of Service attacks
Trending in the dark—what does addresses, the attacker makes it difficult for the known as backscatter. If an attacker randomly
malicious traffic look like? victim’s system to distinguish between the spoofed selects an IP address in our darknet range, and the
As we discussed in the previous section, one of the packets and legitimate packets from real users. The victim responds, we’ll collect that response. By
many data resources that IBM security analysts use victim system will respond to a certain percentage studying these responses we can learn things
to determine trending is the darknet, also known as of these spoofed packets. These responses are about Denial of Service activity on the Internet.
a black-hole network. A darknet is a large range of
IP addresses on the Internet that have never had
Backscatter
any services running on them. Our darknet has an
aperture of 25,600 addresses. Generally speaking, 50,000,000
there is no legitimate reason why computers on the 45,000,000
Internet would send packets to addresses in this
range, but in fact they do. Often, traffic into this 40,000,000
network is associated with malicious activity. This
35,000,000
space is continuously monitored and all incoming
traffic is captured in its entirety and stored for 30,000,000
analysis and long-term archiving.
25,000,000
Spoofed Denial of Service attacks
20,000,000
Looking at the data over the past several years, a
couple of patterns begin to emerge. The first trend 15,000,000
is the gradual rise in backscatter activity (Figure 11).
Backscatter is actually a side effect of a spoofed 10,000,000
Denial of Service (DoS) attack. Attackers launching
5,000,000
Denial of Service attacks on the Internet will often
put incorrect source addresses in the packets they 0
are flooding at their victim. This is known as
11/09
1/10
3/10
5/10
7/10
9/10
11/10
1/09
3/09
7/09
9/09
1/07
3/07
5/07
11/07
1/08
3/08
7/08
9/08
11/08
7/07
9/07
5/09
5/08
spoofing. By spoofing randomly selected source
Total Trend Line
Figure 11: Backscatter 24

Section I > Trending in the dark—what does malicious traffic look like? > Spoofed Denial of Service attacks
In the X-Force darknet, each SYN-ACK backscatter since 2007. However, backscatter is subject to a another group. This “shelling match” between
packet received is an indicator that an attacker sent high degree of variability due to the nature of what warring camps can result in a sudden increase in
a spoofed packet to a well-known service port on is being collected and what is occurring. Some backscatter traffic and backscatter source
the machine under attack spoofed from one of intense periods of backscatter are the result of addresses. It generally ceases as suddenly as it
X-Force darknet addresses. While there has been a internecine warfare within and between various began. This type of activity most likely contributed
gradual increase in backscatter activity since 2007, attacker camps. During this warfare, one group to the dramatic spikes in February 2007 and
there was a large jump year-over-year between attempts to block or take over the resources of December 2009 as shown in Figure 11 on page 24.
2008 and 2009. Part of this increase is due to a
significant spike in activity in 2009—the largest in
the three and half year period. This trend of higher
than previous year averages continues in 2010. At Backscatter - Averages
the close of Q2, the average count for the first half 20,000,000
of 2010 is slightly higher than the total average for
18,000,000
2009, just over 16.5 million. At the close of the year
2010 we see that this number has now jumped to 16,000,000
over 18 million. Figure 12 indicates the increase in
volume from 2007 through 2010 of spoofed Denial 14,000,000
of Service attacks on the Internet.
12,000,000
What can we deduce from this gradual rise in
10,000,000
backscatter data and, in some instances, large
jumps of backscatter activity? Since the majority of 8,000,000
the backscatter data results from Denial of Service
(DoS) attacks, we can speculate that there has 6,000,000
been a steady increase in spoofed DoS attacks
4,000,000
2,000,000
0
2007 2008 2009 2010
Figure 12: Backscatter – Averages 25

Section I > Trending in the dark—what does malicious traffic look like? > Targets of Denial of Service attacks
Targets of Denial of Service attacks Top Backscatter Generating Countries

The nature of a spoofed Denial of Service attack 35,000,000
makes it difficult to determine the attacker. The
attacker fabricates origins for the connections to the 30,000,000
victim’s IP address. These fabricated connections 25,000,000
can in turn come from a multitude of different
machines. When looking at backscatter in the 20,000,000
X-Force darknet, it is clear that the origins of the
15,000,000
attack are spoofed, but the target of the attack is
known. Examining the sources of the backscatter 10,000,000
provides information on the targets of spoofed
5,000,000
Denial of Service attacks. Figure 13 shows the top
backscatter-generating countries for the second half 0
of 2010 as calculated using the WorldIP database 7/2010 8/2010 9/2010 10/2010 11/2010 12/2010
that maps addresses to countries.
US China Turkey Singapore Taiwan Other
There is a fairly common trend in the data. The
United States is by far the largest generator, China is
Figure 13: Top Backscatter Generating Countries
second, and Turkey is third. The United States and
China have the first and second largest counts of IP
addresses so their ranking as backscatter
generators isn’t surprising. If any IP address is as
likely to be a target as any other then one would
expect to see Japan, Germany, South Korea, or the
UK in the top three. This assumption is clearly
wrong. Further analysis and correlation with other
data may help shed light on the matter.
For more discussion about brute force attacks and

the information reported earlier in the year, please
refer to the 2010 Mid-Year Trend & Risk report
located on our web page under report archives. 26
Section I > Worms of yesteryear: Where are they now?
Worms of yesteryear: IBM’s Managed Security Service tracks the Blaster appeared in August of 2003 and rapidly
Where are they now? malicious activity seen on its customer’s networks spread. This worm propagated by exploiting a
The ongoing war against the threat of computer and thus affords a window into the activity of these buffer overflow in the Remote Procedure Call (RPC)
worms is cyclic. A new invader appears, after the worms of yesteryear. The following list gives an interface and the Distributed Component Object
battles to contain the infection and the initial overview of five of the most recent worms that Model (DCOM) interface that had been patched a
outbreak appears won, it falls off the collective spread entirely or partly by exploiting operating month earlier. The worm payload would install an
radar as years pass and the next invaders appear. system vulnerabilities. All of these worms targeted auto-starting executable that would continue trying
software, usually operating systems, by Microsoft. to propagate and trigger a Denial of Service against
Worms propagate by a number of methods such as Microsoft’s update site at a specific time.
malicious email attachments, open or weakly SQL Slammer first appeared in late January of
protected shares and network accessible software 2003, generating such a deluge of traffic that it Sasser appeared at the end of August in 2004.
vulnerabilities. A number of prominent worms have brought down numerous critical resources and It propagated by exploiting a vulnerability in the
appeared over the last seven years but those that noticeably slowed the Internet. Its single UDP Local Security Authority Subsystem Service
spread via exploitation of network accessible packet payload targeted a vulnerability in Microsoft (LSASS), which is a process in Microsoft Windows
vulnerabilities tend to be the most virulent. They can SQL Server that had been patched previously in operating systems that is responsible for enforcing
spread across networks from machine to machine July. The compromised host would then loop, the security policy on the system. LSASS patched a
without a user interceding to view an email or open spamming copies of itself to random IP addresses, few weeks previously. Once infected, a machine
a file. The autonomous spreading of these worms DoSing (Denial of Service) itself and sending out a would download and install an auto-starting
can lead to high infection rates and frequently, large amount of traffic. executable which would scan and attempt to infect
disastrous side effects occur of machines crashing other machines. The worm itself wasn’t malicious
from unreliable exploitation and potentially crippling but a side effect of its scanning caused crashes
network utilization for virulent worms. and reboots in servers and desktops which had
severe consequences for many companies.
27
The Zotob worm appeared in mid-August of 2005.

It propagated by exploiting a buffer overflow in the Year Worm Vulnerability IBM Signature MS Bulletin
Microsoft Plug and Play service that was patched
earlier in the month. A side effect of this propagation
was crashing and reboots of machines due to the 2003 SQL Slammer CVE-2002-0649 SQL_SSRP_StackBo MS02-039
exploit. Infected machines would download and
install an executable to continue propagation and
2003 Blaster CVE-2003-0352 MSRPC_RemoteActivate_Bo MS03-026
install a backdoor to phone back to an Internet Relay
Chat (IRC) channel for further instructions.
2004 Sasser CVE-2003-0533 MSRPC_LSASS_Bo MS04-011
Conficker was detected in early November of 2008.
Propagation was via a vulnerability in the server service
of all supported Microsft operating systems at the 2005 Zotob CVE-2005-1983 PlugAndPlay_BO MS05-039
time. Later variants added additional vectors such as
weak SMB passwords and infection of USB devices.
Once compromised, the infected machine would 2008 Conficker CVE-2008-4250 MSRPC_Srvsvc_Path_Bo MS08-067
attach itself to a botnet awaiting further commands.
Most variants of this worm also performed controlled
Table 2: Top Worms 2003 - 2008
scanning and infection of further hosts.
28
Figure 14 breaks down the alert activity by worm. Worm Exploitation Averages 2010
For clarity, the alert associated with the worm 2,500,000
activity has been renamed for the worm. In most
cases this network activity is based on detected
exploitation by the worms but this is a tricky 2,000,000
endeavor for a number of reasons. For one, the
alert is not necessarily an indication of an attempted 1,500,000
propagation by a worm, alerts can be due to a
security audit or an exploitation attempt by
something else entirely. Another issue is that worms 1,000,000
have different propagation rates. Conficker
regulates its propagation in an attempt to avoid 500,000
overt detection while SQL Slammer can spam
hundreds of exploitation attempts a second. Due to
0
the number of ways that Conficker variants can
Jan
Feb
Mar
April
May
June
July
Aug
Oct
Nov
Dec
Sept
spread, counts of peer to peer activity were used.
SQL Slammer8 has by far the largest number of

exploitation attempts. Even though seven years
have passed in which time to remediate the worm, SQL Slammer Conficker Blaster Sasser
it remains extremely noisy. The large dip in activity
between July and August is due to remediation in a
Figure 14: Worm Exploitation Averages 2010
single network. Ever afterwards, Slammer counts
overshadow all others.
8 SQL slammer traffic on the Internet significantly declined in March 2011 shortly before publication of this report.
For more information on this topic, please see the Frequency-X blog. (http://blogs.iss.net/index.html) 29
Figure 15 shows the same monthly averages with Non-Slammer Averages 2010
SQL Slammer removed. Conficker traffic is the next 200,000
highest. This is not surprising as it is the most
180,000
recent of the studied worms and also known to be
extremely widespread. There is a noticeable decline 160,000
in activity over the year, likely attributed to infected 140,000
nodes being cleaned or brought offline. Blaster and 120,000
Sasser are still showing activity while Zotob’s counts 100,000
were so low that it was removed from the figure. This
80,000
discrepancy may be due to the fact that Blaster and
Sasser would affect both Windows 2000 and XP 60,000
and they came at an earlier time while Zotob only 40,000
affected Windows 2000 and came out in 2005. 20,000
0
It is interesting to note, that the worms exploiting
Jan
Feb
Mar
April
May
June
July
Aug
Oct
Nov
Dec
Sept
vulnerabilities patched over seven years ago still
show noticeable activity. The activity for all the
worms is unlikely to grow significantly as any new
machine brought online should not be vulnerable to
the exploits they spread by. It seems inevitable that Conficker Blaster Sasser
the activity from these worms will eventually die out
as old infected machines are replaced but they do Figure 15: Non-Slammer Averages 2010
show a remarkable tenacity.
New worms will always be on the horizon. As the

latest invader is brought under control and gradually
driven out, it will likely never be fully ousted. There
will almost certainly be a few survivors holding out
in the dark corners of our networks.
30
Section I > Web content trends > Analysis methodology
Web content trends Table 3 below lists the IBM web filter categories that Analysis methodology
This section summarizes the amount and distribution correspond with these types of sites. X-Force captures information about the distribution
of “bad” web content that is typically unwanted by of content on the Internet by counting the hosts
businesses based on social principles and corporate This section provides analysis for: categorized in the IBM Security Solutions web filter
policy. Unwanted or “bad” Internet content is • Percent and distribution of web content that is database. Counting hosts is a method for determining
associated with three types of websites: adult, social considered bad, unwanted, or undesirable content distribution and generally provides a realistic
deviance, and criminal. • Increase in the amount of anonymous proxies assessment. Results may differ when using other
• Malware URLs: Hosting countries and linkage methodologies such as counting web pages and
The web filter categories are defined in detail at: sub-pages.
http://www-935.ibm.com/services/us/index.
wss/detail/iss/a1029077?cntxt=a1027244 The IBM Content data center constantly reviews and
analyzes new web content data. The IBM Content
data center analyzes 150 million new web pages and
Website Type Description & Web Filter Category images each month and has analyzed 14 billion web
Adult Pornography pages and images since 1999!
Erotic / Sex
Social Deviance Political Extreme / Hate / Discrimination The IBM Web Filter Database has 68 filter categories
Sects and 67 million entries with 150,000 new or updated
Criminal Anonymous Proxies entries added each day.
Computer Crime / Hacking
Illegal Activities
Illegal Drugs
Malware
Violence / Extreme
Warez / Software Piracy
Table 3: Web filter categories associated with unwanted web content
31
Section I > Web content trends > Percentage of unwanted Internet content
Percentage of unwanted Internet content Content Distribution of the Internet

Approximately seven percent of the Internet 2010
currently contains unwanted content such as
pornographic or criminal websites.
Increase of anonymous proxies

As the Internet becomes a more integrated part of our
lives—not only at home, but at work and at school—
Adult: Criminal
organizations responsible for maintaining acceptable 0.245%
6.635%
environments increasingly find the need to control
where people can browse in these public settings. Other:
0.248%
93.117%
One such control is a content filtering system that

prevents access to unacceptable or inappropriate
websites. Some individuals attempt to use an
anonymous proxy (also known as web proxies) to
circumvent web filtering technologies. Social
Deviance
0.003%
Figure 16: Content Distribution of the Internet – 2010
32
Web proxies allow users to enter a URL on a web Volume Increases of Anonymous Proxy Websites
form instead of directly visiting the target website. H2-2007 to H2-2010
Using the proxy hides the target URL from a web 600%
filter. If the web filter is not set up to monitor or block
anonymous proxies, then this activity (which would 500%
have normally been stopped) can bypass the filter
and allow the user to reach the disallowed website. 400%
The growth in volume of anonymous proxy websites 300%

reflects this trend.
200%
In the past three years, anonymous proxies have
100%
steadily increased, more than quintupling in number.
Anonymous proxies are a critical type of website to
0%
track, because of the ease at which proxies allow H2-2007 H1-2008 H2-2008 H1-2009 H2-2009 H1-2010 H2-2010
people to hide potentially malicious intent.
Figure 17: Volume Increases of Anonymous Proxy Websites – H2-2007 to H2-2010
33
Top Level Domains of Anonymous Proxies Top Level Domains of Newly-Registered Anonymous Proxy Websites
Figure 18 illustrates the Top Level Domains (TLDs) Q1-2006 to Q4-2010
of the newly-registered anonymous proxies.
70%
In 2006, more than 60 percent of all newly- 60%

registered anonymous proxies were .com domains,
but since the middle of 2007, .info has been at the 50%
top until the beginning of 2010 (while .com was
40%
runner-up most of the time).
30%
But why is .info no longer in the prime position? It
seemed to be a proven TLD for anonymous proxies 20%
for years. A reason could be that .info, similar to .com,
10%
is running out of names. Additionally, the question
arises why anonymous proxies are now provided on 0%
.cc and .tk top level domains. These are the Domains
9
8
07
of Cocos (Keeling) Islands (.cc), an Australian territory,
06
09
08
10
07
10
0
0
0
20
20
20
20
20
20
20
20
20
20
and Tokelau (.tk), a territory of New Zealand. Nearly all
3-
3-
1-
3-
3-
3-
1-
1-
1-
1-
Q
Q
Q
Q
Q
Q
Q
Q
.cc anonymous proxy websites are registered on the

domain co.cc. It is free of charge to register a domain
anything.co.cc (see http://www.co.cc/?lang=en). info com cn net org
The same is true for .tk. (see http://www.dot.tk/). li ch cc tk
Thus, it is both cheap and attractive to install new
anonymous proxies on .co.cc or .tk. Figure 18: Top Level Domains of Newly-Registered Anonymous Proxy Websites – Q1-2006 to Q4-2010
Additional trends:
• At the end of 2009, .cc (Cocos (Keeling) Islands)

started to increase significantly and even reached
the number one position in the second quarter of
2010. Nevertheless .cc went out of vogue by the
end of 2010.
34
• In the second quarter of 2010, another new star in Newly-Registered Anonymous Proxy Websites
proxy heaven, .tk (Tokelau), reached about 23 United States Hosted vs. Not United States Hosted
percent of new anonymous proxies. It dominated Q1-2006 to Q4-2010
the rest of the year by acquiring nearly 30 percent
100%
in the third quarter and more than 56 percent in the
fourth quarter of 2010. 90%
• During that same time period, .info decreased 80%
dramatically and fell below 10 percent for the first 70%
time by the end of 2010. 60%
• In the first quarter of 2010, even .com fell 50%
significantly below 20 percent for the first time, 40%
recovering to 26 percent and then 19 percent in
30%
the third and the fourth quarters of 2010.
20%
It will be interesting to see whether .tk has a similar 10%
destiny as .co.cc—being the star of anonymous 0%
proxies for a year and a half before declining.
08
09
07
06
09
08
10
07
10
0
20
20
20
20
20
20
20
20
20
20
3-
3-
1-
3-
3-
3-
1-
1-
1-
1-
Country hosts of anonymous proxy websites
Q
Q
Q
Q
Q
Q
Q
Q
For anonymous proxy hosting countries, the United

States has held the top position for years. More
than 70 percent of all newly registered anonymous USA Others
proxies were hosted in the U.S. for the years
2006-2009. In the third quarter of 2010 they fell Figure 19: Newly-Registered Anonymous Proxy Websites United States Hosted vs.
below 70 percent for the first time in more than four Not United States Hosted – Q1-2006 to Q4-2010
years, but recovered to nearly 72 percent by the
end of 2010.
35
It is worth looking at the remaining 30 percent of all Non United States Newly-Registered Anonymous Proxy Websites
newly registered anonymous proxies in 2010. This Q1-2006 to Q4-2010
remainder is dominated by UK (9 percent in the 18%
third quarter of 2010), Canada (6.4 percent in the
16%
third quarter of 2010), and Netherlands (5.8 percent
in the third quarter of 2010). Thus, those three 14%
countries made up more than 20 percent in the 12%
third quarter of 2010. All other countries host less 10%
than 4.5 percent at the time of press in 2010.
8%
6%
4%
2%
0%
09
08
07
06
09
08
10
07
10
0
20
20
20
20
20
20
20
20
20
20
3-
3-
1-
3-
3-
3-
1-
1-
1-
1-
Q
Q
Q
Q
Q
Q
Q
Q
Germany Canada United Kingdom Netherlands

Australia China France
Figure 20: Non United States Newly-Registered Anonymous Proxy Websites – Q1-2006 to Q4-2010
36
Section I > Web content trends > Malicious websites
Malicious websites Geographical location of malicious web links malicious URL hosting countries, claiming 7.9
This section discusses the countries responsible for The United States continues to reign as the top percent (as shown in Figure 21).
hosting the malicious links along with the types of hosting country for malicious links. More than one
websites that most often link back to these third of all malware links are hosted in the U.S. The second-tier countries have also shifted, and,
malicious websites. Exploits from Malicious While China was on top two years ago, it is runner- most significantly, many more countries seem to be
websites discusses the web exploit toolkits involved up in 2010, hosting 8.5 percent—only 0.2 percent jumping into the game.
in the majority of these malicious websites. more than France. Romania is new within these top
Countries Hosting the Most Malicious URLS Second-Tier Countries Hosting Malicious URLs
2006-2010 2006-2010
90% 10%
80% 9%
8%
70%
7%
60% 6%
50% 5%
4%
40%
3%
30% 2%
20% 1%
0%
10%
2006 2007 2008 2009 2010
0%
2006 2007 2008 2009 2010
Germany Russia Israel Canada
South Korea Netherlands United Kingdom
USA China France Romania Others
Figure 22: Second-Tier Countries Hosting Malicious URLs – 2006-2010

Figure 21: Countries Hosting the Most Malicious URLS – 2006-2010
37
Good websites with bad links in hopes of finding a loop-hole (like a vulnerability or The chart below lists the most common types of
As described in Web Application Vulnerabilities an area that allows user-supplied content) in which websites that host at least one link that points back
and Common Domains in URL Spam, attackers they can incorporate malicious links in hopes of to a known malicious website.
are focusing more and more on using the good name compromising an unsuspecting victim.
of trusted websites to lower the guard of end users
and attempt to obfuscate their attacks with protection
technologies. The use of malicious web content is Top Website Categories Containing at Least One Malicious Link
no different. The following analysis provides a H2-2010
glimpse into the types of websites that most
frequently contain links to known, malicious links. Pornography or erotic/sex sites
Personal homepages or
Some of the top categories might not be surprising. For communication services sites
example, one might expect pornography and gambling Gambling/lottery sites
to top the list. Indeed, together they own more than
Search engines/web
30 percent of all malicious links. However, the second catalogs/portals sites
-tier candidates fall into the more “trusted” category.
Education sites
Blogs, bulletin boards, search engines, personal Blogs/bulletin boards sites

websites, shopping sites, education, online
News/magazines sites
magazines, and news sites fall into this second-tier
“trusted” category. Many of these websites allow Shopping sites
users to upload content or design their own
website, such as personal content on a university’s Computer games sites
website or comments about a purchase on a Anonymous proxies sites
shopping website. It is unlikely that these types of
websites are intentionally hosting malicious links. Other sites
The distribution is probably more representative of
0% 5% 10% 15% 20% 25% 30% 35% 40%
the types of websites that attackers like to frequent
Figure 23: Top Website Categories Containing at Least One Malicious Link – H2-2010
38
When comparing this data with the data of the Top Website Categories Containing at Least One Malicious Link:
previous years, interesting trends appear. Types of Sites on the Incline
Particularly in the first half of 2010, professional H1-2009 to H2-2010
“bad” websites like pornography or gambling
35%
websites have increased their links to malware,
making it appear more likely that “professionals” are 30%
improving their efforts to systematically distribute 25%
their malware. However, in the second term of 2010
they declined again, but both end in a percentage 20%
above the levels of 2009. 15%
Educational sites such as university websites have 10%

also seen increases in malware links since 2009. 5%
The same is true for Blogs and bulletin boards until
mid-2010. Then they significantly decreased and fell 0%
below 5 percent for the first time in more than a H1-2009 H2-2009 H1-2010 H2-2010
year. Moreover, we noticed increases for computer
games and anonymous proxy sites, but on a Pornography or Computer games sites
lesser level. erotic/sex sites Anonymous proxies sites
Blogs/bulletin boards sites Education sites
The only major category that did not decrease
Gambling/lottery sites
significantly in the second half of 2010 was
gambling sites.
Figure 24: Top Website Categories Containing at Least One Malicious Link: Types of Sites on
the Incline – H1-2009 to H2-2010
39
Personal homepages are no longer the most Top Website Categories Containing at Least One Malicious Link:
prevalent category that host at least one malicious Types of Sites on the Decline
link. Personal homepages have improved—they H1-2009 to H2-2010
now host less malicious links—compared to the
first half of 2009. One reason may be that personal 60%
homepages are more out of style in favor of web
2.0 applications such as profiles in social or 50%
business networks. Search engines, portals,
shopping sites, and news sites have also improved 40%
or stayed on a low level. These traditional legitimate
interactive sites have been used to exchange 30%
information and opinions for years. Thus, it is likely
that providers of those services have increased their 20%
efforts in IT security.
10%
0%
H1-2009 H2-2009 H1-2010 H2-2010
Personal homepages or Shopping sites

communication services sites News/magazines sites
Search Engines/
web/catalogs/portals sites
Figure 25: Top Website Categories Containing at Least One Malicious Link: Types of Sites on
the Decline – H1-2009 to H2-2010
40
Until now we have not consider the number of Top Website Categories Containing Ten or More Malicious Links
malicious links placed on a website. The difference H2-2010
might be:
Pornography or erotic/sex sites
• When hosting only one or two malicious links on a
site, the owner of the site might not understand or
Gambling/lottery sites
know that the link is bad – there is no ill intent.
• When placing ten or more links on a site, then this
Search engines/web
is done systematically and intentionally to get
catalogs/portals sites
visitors clicking on bad links. The goal of the owner
might be to enjoy a financial advantage from the
Personal homepages or
compromises. communication services sites
Out of the categories of websites that host 10 or

Blogs/bulletin boards sites
more of these links, pornography accounts for
nearly 30 percent and gambling accounts for nearly
Anonymous proxies sites
29 percent.
News/magazines sites
Compared to the data six months ago, the values in
most categories have stayed flat or slightly
Computer games sites
decreased but gambling increased by nearly one
percent. Against the background of 0.6 percent of
Education sites
the adult population having problem gambling
issues (see http://en.wikipedia.org/wiki/
Shopping sites
Gambling_addiction#Prevalence), gambling sites
are a popular target for malware distributors. Note
Other sites
also that Personal Homepages and Communication
Services increased by 1.7 percent and Educational
0% 5% 10% 15% 20% 25% 30% 35%
sites increased by 0.6 percent.
Figure 26: Top Website Categories Containing Ten or More Malicious Links – H2-2010
41
Section I > Spammers focus on content rather than volume > Major content trends in spam for 2010
Spammers focus on content rather Major content trends in spam for 2010 in technical spam content was the constant level of
than volume After the last major threats of image-based and PDF HTML-based spam (in most cases a bit more than 80
The IBM spam and URL filter database provides a spam in 2007, we did not see major changes in the percent) and plain-text spam (mostly 10-15 percent).
world-encompassing view of spam and phishing content of the spams in 2008 and 2009, apart from
attacks. With millions of email addresses being another short-period threat of image spams in the first In 2010 there were major changes in the technical
actively monitored, the content team has identified term of 2009. One characteristic for the low changes content of spam. To see these trends at a glance,
numerous advances in the spam and phishing see Figure 27.
technologies attackers use.
Major Content Trends in Spam
Currently, the spam filter database contains more 2010 per week
than 40 million relevant spam signatures. Each 14 6%
piece of spam is broken into several logical parts
(sentences, paragraphs, etc.). A unique 128-bit 12
Kilobyte/Number of URLs per Spam 5%
signature is computed for each part and for millions
of spam URLs. Each day there are approximately 10
one million new, updated, or deleted signatures for 4%
the spam filter database. 8
3%
This section addresses the following topics: 6
• Major content trends in spam 2010 2%

4
• Most popular domains and top level domains used
in spam 2 1%
• Spam country9 of origin trends, including spam
web pages (URLs) 0 0%
• Most popular subject lines of spam
49
29
1
45
5
33
13
17
21
37
41
25
k
k
k
ee
ee
ee
k
k
k
k
k
ee
ee
ee
ee
ee
ee
ee
ee
ee
ee
W
W
W
W
W
W
W
W
9 The statistics in this report for spam, phishing, and URLs use
the IP-to-Country Database provided by WebHosting.Info
(http://www.webhosting.info), available from http://ip-to- Average byte size of spam Percentage ZIP spam
country.webhosting.info. The geographical distribution was Average number of URLs per spam Percentage HTML attachment spam
determined by requesting the IP addresses of the hosts (in the
case of the content distribution) or of the sending mail server (in
the case of spam and phishing) to the IP-to-Country Database. Figure 27: Major Content Trends in Spam – 2010 per week 42
Let’s have a closer look at the trends and the • August—Spam with malicious ZIP attachments:
characteristics: At the beginning of August, spammers began Zeus Trojan
sending spam threats with ZIP attachments. We Zeus is a very common Trojan that’s generated
• March-August: Random text spam combined looked into these messages, and each ZIP file with a kit that anyone can purchase online.
with random URLs, significantly increased the contained a single EXE file that was malicious. There are many different individuals and groups
average byte size of spam. In previous years the Spammers used different kinds of malware, e.g. that have Zeus botnets set up. There are a lot
average byte size of spam was directly dependent variants of the Zeus Trojan or a copy of the of ways it gets spread, but the operators of this
on the percentage of image-based spam. But in Bredolab downloader (see sidebars). More details particular botnet are growing it by sending out
2010 the percentage of image spam was flat and on these spam threats with ZIP attachments can emails with ZIP file attachments. The goal of
below two percent (in most cases below one be found at http://blogs.iss.net/archive/ Zeus botnets is usually to steal personal
percent). When looking at these larger spams one ZIPMalwareSpam.html. IBM Proventia customers information, and the type of information stolen
can see large text fragments randomly chosen can use the Email_Zip_Executable_Content is commonly online banking data that criminals
from the Internet, complemented by random URLs signature to detect threats like these. The can use to access bank accounts to transfer
(syntactically correct URLs build from random spammers used typical methods to attract the money. For more information about the Zeus
characters or words, but many of them do not user’s attention by using subjects such as: botnet see Trojan Bot networks in the section
exist in the Internet). Random text is a very old –– Your Flight Ticket “IBM Managed Security Services—
technique of the spammers to make spam look –– Financial Summary A global threat landscape”.
more legitimate. However, recent anti-spam –– Statement Notification
techniques do not have any problems with it. So –– Financials
why did spammers re-activate this old approach? –– FW: Car & Car loan
Maybe they hoped that those masses of text –– Employee Orientation report
would confuse Bayesian classifiers, particularly
self-trained Bayesian classifiers, which are used
in a non-business context; hence, these spam
attacks might be targeted to these
non-business users.
43
• September—Spam with HTML attachments: • September-November—Random URL spam:

There are some similarities between the ZIP During this time period, spammers did not use Bredolab downloader
attachment spam emails of the month before and random text but instead used random URLs This Trojan downloads a rogue antivirus
the HTML attachment spam. In both cases the extensively. This resulted in more than 12 program called SecurityTool that pretends to
user’s computer gets infected when clicking on the (syntactically correct but otherwise useless and find viruses on your PC when none exist.
attachment. Furthermore, in the HTML attachment random) URLs per spam on an average during the
spam, the user’s attention is attracted in the same beginning weeks of this time period. In the
way as in the ZIP attachment spam by something following weeks, we recognized more than six
in the email text body such as: URLs per spam, which is still above the normal
–– Please see attached invoice for Stockton levels of 2-4 URLs per spam on an average.
floor project
–– More details are in the attached invitation • December—Increased average byte size of
–– See attached for breakdown—the $40 HOA spam again: This time, this is a result of the drop
will not be included in payment do deduct of the spam volume by 70 percent (see section
from total which = $1095/mo “Spammers on holiday at the end of the year”
–– Please print out the invoice copy attached page 46), that affected particularly small spam.
and collect the package at our office
–– Attached is a copy of the deposit received for Against the trends of previous years—wherein
your records spammers made very few changes in the technical
–– Here are the signed documents content of spam throughout the year, in 2010,
–– memo on image secrecy (attached) spammers made a continuous effort to change the
–– Enclosed is my CV for your consideration technical contents regularly. In the next section, we will
–– You will find the resume attached to this email discuss an associated factor—the volume of spam.
–– Attached you will find the fall daily tour
schedule for your review
44
Section I > Spammers focus on content rather than volume > Spam volume > Conclusions about spam volume and content
Spam volume –– the growth of Backscatter Activity in 2010 - • Are the new operating systems more secure and
While we recognized significant increases of the spam see section “Spoofed Denial of prevent a further increase of the levels?
volume year over year until 2009, in 2010 there were a Service attacks” • Do even “spamming companies” suffer from the war
few months with ups and downs in the volume of –– the rise of the Vulnerability Disclosures in for talent, hence, they have recruitment problems?
spam seen over the year. However, the overall trends 2010 - see section “2010 - A record
stayed flat, and we saw less volume at the end of the setting year” It is very unlikely that the spam business has become
year in comparison to the beginning of 2010. • Is the increase of spam messages only achieved unprofitable. One scenario could be that spam volume
within internal social network messaging systems stays flat but the kinds of spam change more
Conclusions about spam and other Web 2.0 applications? frequently to circumvent spam filters with new types of
• Are spammers cautious in efforts with increasing the spam that are more difficult to detect.
volume and content
Why are spammers making an effort to change the levels too much because the more similar spam
messages they produce, the easier they can be Maybe there will be more experiments with other
technical content of spam more often than in previous
detected and blocked by perfected spam filters? attachment types? We tallied the most popular file
years but are no longer focusing on increasing the
That would mean, they assume that they have types, and there is one file type becoming more and
overall volume of spam? Here we ponder a few
reached an optimum concerning the spam volume. more popular – Open Office documents. When do
presumptions about these possible trends. Some
spammers use those attachments?
trends might be more plausible than others.
• Perhaps in recent years there was a linear connection Changes in Spam Volume
between the number of spam messages and the April 2008 to December 2010
profit reached by spam. Is this connection lost? 250%

• Is the spam market saturated? Will we even see a
200%
decrease of spam volume in the upcoming years?
• Since there is only one single point to combat
150%
current spam - when receiving them - for the
companies (or the users), did the bad guys change 100%
their focus to other - more distributed - areas that
are more complicated to take countermeasures? 50%
This assumption is strengthened by:
0%
–– the increase of Botnet Trojan Activity in 2010 -
see section “Trojan Bot networks”
8
10
10
10
10 0
12 0
0
1
01
01
0
0
20
20
20
20
20
20
20
20
20
20
20
20
20
20
20
/2
/2
2/
4/
6/
8/
4/
6/
8/
2/
4/
6/
8/
/
–– the surge of Obfuscation Activity in 2010 -
12
12
10
10
see section “Obfuscation” Figure 28: Changes in Spam Volume – April 2008 to December 2010 45
Section I > Spammers focus on content rather than volume > Spammers on holiday at the end of the year
Spammers on holiday at Spam Volume per Day

the end of the year December 9th, 2010 until January 11th, 2011
One week before year’s end, spammers surprised us 160%
by sending out 70 percent less spam than the weeks
140%
before; this period lasted about two and a half weeks.
After the Christmas holiday season, spam levels 120%
returned to the same level as before Christmas. 100%
When looking at the reductions of the spam volume 80%

per country there were some countries, such the 60%
U.S., Canada, and UK, with a decline of more than
90 percent. More about the declines per country 40%
and some more details can be found on http:// 20%
blogs.iss.net/archive/2011spambotdecline.html.
0% 10
10
10
/2 10
/2 10
/2 10
1/ 10
1/ 11
1/ 11
1/ 11
1/ 011
1
01
12 201
12 201
12 201
12 201
12 201
20
20
20
20
20
20
20
/2
2
2/
4/
6/
8/
1/
3/
5/
7/
9/
1/
3/
5/
7/
9/
1/
9/
10
/1
/1
/3
/2
/1
/1
/1
/2
/
12
12
12
12
12
12
12
Figure 29: Spam Volume per Day – December 9th, 2010 until January 11th, 2011
46
Section I > Spammers focus on content rather than volume > Regional spam volume per day of the week
Regional spam volume days; on Saturday and Sunday, they only send out Assuming that spammers prefer not to work
per day of the week about five percent each day. Their strongest days are weekends, it appears that spam in the English,
Another approach for looking at the spam volume is Tuesday, Wednesday, and Thursday, when they French, and Spanish languages is sent out
checking the spam volume per day of the week. If process about 20 percent of their weekly amount completely automatically, retaining its typical volume
we received equal volumes each day, then we each day. The patterns are similar for Portuguese on weekends. Contrarily, Russian and Portuguese
would receive 14.3 percent of the weekly spam spam. Their strongest days are Tuesday to Thursday, spam requires more manual work, resulting in a
volume per day. When looking at spam written in and their weakest days are Saturday and Sunday. significant drop at the weekends.
English, French, or Spanish, this appears more or
less the case.
English, French, Spanish, Russian, and Portuguese Spam Volume
English spam is distributed very consistently over 2010 per Day of the Week
the week days. The days with the least amount of 25%
spam are Wednesday (14.0 percent) and Sunday
(13.7 percent); the days of the week with the most
spam are Tuesday (14.7 percent) and Friday (14.8 20%
percent). The most French spam is received on
Thursday (15.7 percent) and Friday (15.8 percent).
The greatest spam day of the week for Spanish 15%
spam is Monday, when they process 18 percent of
the weekly amount of their spam. However, the
10%
difference between the other week day amounts for
French and Spanish is rather low.
5%
The situation is different for Russian and Portuguese
spam. On weekends, we receive much less spam
written in these two languages. Almost 90 percent of 0%
spam in the Russian language is sent out on week Monday Tuesday Wednesday Thursday Friday Saturday Sunday
English French Spanish Russian Portuguese
Figure 30: English, French, Spanish, Russian, and Portuguese Spam Volume – 2010 per Day of the Week 47
Section I > Spammers focus on content rather than volume > Common domains in URL spam
Common domains in URL spam • Internet service providers (blue): Used by • Official websites of Pfizer and Rolex (yellow):
The vast majority of spam, more than 90 percent, is spammers in recent years to make look their From September 2010 on, spammers used the
still classified as URL spam—spam messages that spams appear trustworthy. official websites of Pfizer (pfizer.com,
include URLs that a person clicks to view the spam • Image-hosting websites (green): Also used by pfizerhelpfulanswers.com, viagra.com) and Rolex
contents. It is worthwhile to take a closer look at spammers for several years. Spammers like to vary (rolex.com). Obviously, spammers include in their
the most frequently used domain names in URL between well known image-hosters like flickr.com strategies that most spam filters do not use simple
spam. The table on the following page shows the and imageshack.us and many other small and keyword search anymore and even assume that
top 10 domains per month throughout 2010, with medium-sized image-hosting websites. URLs from pfizer.com or rolex.com make their
some key domains highlighted. • Random word domains (orange): From July to messages looking more legitimate.
September 2010 spammers used random words • URL shortening services (purple): From
The majority of those domain names are well-known to “build” URLs. This was done in such a massive September 2010 on, some of these services
and trusted (highlighted in color in the table on page way that the very common words “the”, “of”, made it to the top 10.
49). Not only do these legitimate websites provide a “and”, “in”, “a” even made it to the top ten with the
recognizable (and trustworthy) web link to the end user, “.com” extension. Since then, we have seen The table of domains on the next page became
but spam messages using them may also successfully random domains built from random characters and more multicolored in the second half of 2010. That
evade some anti-spam technology because they now it appears we see random domains built from means that spammers used multiple methods to
only use legitimate links in their spam emails. There random words. present their offers via URLs. This is another
are different types of well-known domains: illustration of the move of spammers from volume to
“content quality,” as mentioned above.
48
Section I > Spammers focus on content rather than volume > Common domains in URL spam
Rank January 2010 February 2010 March 2010 April 2010 May 2010 June 2010
1. flickr.com radikal.ru livefilestore.com livefilestore.com imageshack.us imageshack.us
2. imageshack.us imageshack.us imageboo.com imageshack.us imageshost.ru imageshost.ru
3. radikal.ru livefilestore.com radikal.ru imageshost.ru myimg.de pikucha.ru
4. livefilestore.com flickr.com imageshack.us imgur.com xs.to imgur.com
5. webmd.com live.com googlegroups.com myimg.de imgur.com mytasvir.com
6. picsochka.ru imageboo.com live.com xs.to tinypic.com mojoimage.com
7. live.com capalola.biz akamaitech.net icontact.com livefilestore.com myimg.de
8. superbshore.com feetorder.ru gonestory.com tinypic.com icontact.com twimg.com
9. tumblr.com laughexcite.ru bestanswer.ru live.com googlegroups.com icontact.com
10. fairgreat.com hismouth.ru wrotelike.ru binkyou.net images-amazon.com twitter.com
Rank July 2010 August 2010 September 2010 October 2010 November 2010 December 2010
1. imageshack.us yahoo.com the.com businessinsider.com rolex.com pfizer.com
2. icontact.com the.com of.com migre.me msn.com viagra.com
3. the.com icontact.com msn.com 4freeimagehost.com bit.ly msn.com
4. myimg.de feetspicy.com pfizerhelpfulanswers.com bit.ly pfizer.com rolex.com
5. of.com of.com and.com postimage.org co.cc bit.ly
6. imgur.com ratherwent.com bit.ly imgur.com royalfoote.com product45h.com
7. by.ru and.com in.com pfizer.com royalbelie.com newpfizermed5k.com
8. and.com facebook.com yahoo.com viagra.com royalreleasable.com xmages.net
9. in.com in.com a.com uploadgeek.com luxurystorewatch.com cordfork.com
10. tastymighty.com a.com x-misc.com vipplayerq.com basincook.com onlinepfizersoft2.com
Table 4: Most common domains in URL spam, 2010
49

2011 IBM X-Force Trend and Risk Report

Uploaded by

Copyright:

Available Formats

2011 IBM X-Force Trend and Risk Report

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2011 IBM X-Force Trend and Risk Report

Uploaded by

Copyright:

Available Formats

IBM X-Force® 2010

Trend and Risk Report

IBM Confidential and Internal only—2011 2

Contents > Section I

Contributors 3 Spoofed Denial of Service attacks 24 Phishing volume 57

Section I—Threats 7 Web content trends 31 Phishing—country of origin trends 60

Executive overview 7 Analysis methodology 31 Phishing URLs—country of origin 61

Threats 8 Malicious websites 37 Phishing—most popular subject lines 63

Developing Secure Software 9 Major content trends in spam for 2010 42

Emerging Trends in Security 10 Spam volume 45

IBM Security collaboration 10 Conclusions about spam volume and content 45

SQL injection 13 Common domains in URL spam 48

Obfuscation 15 Common top-level domains in URL spam 51

PDF exploitation 16 Internationalized country code top-level domains:

Contents > Section II, III and IV

Navigating the report

Navigating the report

Section I > Executive overview

Section I > 2010 Highlights > Developing Secure Software

Emerging Trends in Security

1 Massive Mariposa botnet shut down – http://www.net-security.org/secworld.php?id=8962

SQL injection SQL Injection Attacks Monitored by

Figure 2: SQL Injection Attacks Monitored by IBM Managed Security Services

Figure 3 illustrates the significant SQL injection attack SQL_Injection_Declare_Exec Activity

Obfuscation Obfuscation Activity

Obfuscation activity continued to increase during 0

PDF exploitation PDF Activity

Cross-site scripting Cross-Site Scripting Activity

Industry trends Cross-Site Scripting - Industries with Downward Trend

Figure 7: Cross-Site Scripting – Industries with Downward Trend

Figure 8: Cross-Site Scripting – Industries with Downward Trend

Section I > Top high-volume signatures—IBM MSS

Top high-volume signatures— Rank Event Name Trend Line

HTTP_Unix_Passwords SMB_Empty_Password SQL_Injection HTTP_Unix_Passwords SMB_Empty_Password SQL_Injection

spoofing. By spoofing randomly selected source

Total Trend Line

Figure 11: Backscatter 24

Figure 12: Backscatter – Averages 25

Targets of Denial of Service attacks Top Backscatter Generating Countries

For more discussion about brute force attacks and

Section I > Worms of yesteryear: Where are they now?

Section I > Worms of yesteryear: Where are they now?

The Zotob worm appeared in mid-August of 2005.

Section I > Worms of yesteryear: Where are they now?

SQL Slammer8 has by far the largest number of

Section I > Worms of yesteryear: Where are they now?

New worms will always be on the horizon. As the

Section I > Web content trends > Analysis methodology

Table 3: Web filter categories associated with unwanted web content

Percentage of unwanted Internet content Content Distribution of the Internet

Increase of anonymous proxies

One such control is a content filtering system that

Figure 16: Content Distribution of the Internet – 2010

The growth in volume of anonymous proxy websites 300%

In 2006, more than 60 percent of all newly- 60%

and Tokelau (.tk), a territory of New Zealand. Nearly all

.cc anonymous proxy websites are registered on the

• At the end of 2009, .cc (Cocos (Keeling) Islands)

Country hosts of anonymous proxy websites

For anonymous proxy hosting countries, the United