2009 Databreach Report

01000101010101100100111001010100010110000100100101000111010110010100100101001101010101110101001101001110010001010
10010000100010101001001010001010100011001001111010101000101100001000010010100110100001101010111010110010100100001
01001001010001010011010101011101000111010101010101101001000001010000100101011001011001010000110100001001000010010
00110010100100100010101011001010001100100001001010110010001010100010001001011010001010101011001001101010001100101
00100100100101000110010011100100011101000110010011100101001001000010010001100100011101010110010010110101001101000
11001010000010011100100001001010101010001100101101001001010010001110100001101000101010001010100010101010111010000
01010010110100100001010000010110000100010101000010010101000101101001001010010000110101101001001111010101110100011
10101010001000010010100110101000101000111010101000100110101001001010000010101100101000100010100000101100101000100
01010010010010010101001001011001010001010101010001001011010000110100101001010010010100000101100101001000010001010
10100000101011101001011010101010100111101000001010001010100101101001110010101100101010001010110010110100100100001
01001101001101010110100100111001010100010101000100100101010110010010010100101101001101010011010101001001011001010
10011010011100101010101001001010000010100101101000010010100100100101101010001010011010101001101010100010110010100
00110100011101000011010000110101001001001100010100100101001001001001010010010101001001000101010001100100011101011
00101010100010010100101010101000010010101010101100001001000010001010101100101010011010001110100110001000101010110
01010100100101011001001000010010010101100101011000010001000100010101011001010110100100001101001010010010110101011
00101010001001111010100110100111101001001010110000100101001000101010010000100111101011000010001010101011001001101
01010111010010100100001001001110010110100100110101010100010010110101011101011010010001010100011001001111010001100
10000110100111001000010010101110100111001000011010101010101011101001101010110010100011001001001010101010101011001
00001001001011010101110100111001010000010101110101010001011001010011110100010101011001010100010101010001001001010
10010010100100101100101010010010000110100110101001110010101100100011001010110010011000101001001010011010000100100
11100101010001010000010101110101000001000001010011110100001101011010010100000100010101001011010010000100110001000
11001000011010001010100010101010010010100100101011001010111010101100101010101011001010000100101011001001010010100
00010101010101011001010000010011110100000101011001010011010100100101001011010100010101000101001110010100110101011
10101101001000111010010000101101001001011010001000100011101011001010011000100000101000101010001110101011101010000
01001011010001010101001101000111010000110101100101011010010001100101011001001010010001000100110101000101010100000
10100010100101101010011010100110100110001001110010101100101001101010110010100000101010101010110010101100101001001
01011001011001010001010101001001001000010001000101010001010101010101000101100101011001010011010101000101000111010
00101010101100101011101010010010011010101000101010011010110100100011001001110010100000100111001010010010010100100
10010100011101000111010101110100000101001010010011100100111001001010010011000100101101001111010001010101000101001
00001001110010001010101010001010010010100000101010101010001010110010100010001000110010110100101011101000011010110
10010010110101011001001010010001010101100001001100010011010100001101001011010000110101001101001001010001100101010
00100001101010100010100110101010101010100010011000100010001010010010100100100110101001001010010110101000101010100
01001110010010010100111001010000010001110101001001010000010100010101000101011000010100000101010001011010010001000
10100000100000101001001010011110101010001000011010001010101010101000001010110100100011001000101010101110100010001
01000101001100010011000101000001011010010100100100100001011000010011000101100001010001010001110101001101001100010
10010010010100101010001000010010011000101101001010010010010010101001001010110010010010101001101001110010110100100
10010101011101001100010011010101011001011001010000010100010001010110010011110100100001000110010001010101011001001
11001000001010010110100101101000111010011110101001001010010010110000101001101011001010001110101100001010000010101
01010011010101011001000111010000100100111101001101010100100100101001001100010000110101001001000101010001100100001
10100110101010010010100010101011001011000010101000100110101001001010110010100110101001010010010100101011001001000
01011000010011100100001001010100010100110101101001001101010101000100101001000101010001100100101101000110010001110
10010110101010101010010010001100100110001001110010010000101100001010000010010110100001101010111010011000100010101
01100001001101010010010101100101001100010001110101100101001110010011100101001001010111010000010100101101010011010
00101010101110101010001001000010100000100101101000111010110100100101101001011010110000100011101000001010110100100
01010100110001001100010101010101010001000001010110010100001101001001010001010100101101010111010010010101001101001
00001010101010011100100010001001011010001010100101101010111010000010101001001000111010000100101100101011010010001
10010001110100101101000101010100000100101101010001010001110101101001011010010100110101001001001001010011010100011
00100110001000111010010110100000101010010010101000101010101010010010000010100100101001110010100110100111001000111
01000101010001010101010101001101010001010101100001010010010101100100010101000101010011000101101001011000010101000
10010010101001101010101010101110101011001011010010010110100111101011001010011000101010001010000010000100100100001
01101001010111010001010100111101010001010101110100111001011000010011100101000001011000010100000100101101010011010
10011010110000100101001001000010100000100000101001110010000110101011001000110010100000101001001011001010000010100
01000101001001001100010100100100111101000101010101110100010101000010010100010100010101010111010010000101101001010
010010001110100000101 285 MILLION RECORDS WERE COMPROMISED IN 2008. 0101000101101001000100
010001110101010101000 0110100010101001011010
01100010001100101100101001000010110100100101001001110010011100101101001001001010010100101001001000111010011100101
10100101001001010110010000100100111101000011010000010101010101011001010001010101101001000111010010110101000001010
01101001010010110000100101001001001010000010101001101001101010101100100011001010100010001000101011101000110010110
00010000100100100101000100010010000101000101011010010001010101100101001011010001000101001001010100010001000101001
00100100101001111010100000101000001001011010010100101001001010000010010010101001101010011010010110100110101000011
01011010010010100100011001011010010101000100001001010110010000100100101001010101010001110100010101011001010000010
10011100100101001001001010001110100101001010100010001000100001101010000010101000101101001000100010001010100111101
00011101010101010101000100110001011010010100000100010101001011010010000101010001001110010010010100100001010100010
00111010001110101010101001101010101100100011101000010010011110100110101010010010010100100110001000011010100100100
010101000110010100110101011101000110010110100100111101000011010100100100111101001000010001010100000101010101

A study conducted by the Verizon Business RISK Team

2009 Data Breach
Investigations Report

2009 Data Breach Investigations Report
A study conducted by the Verizon Business RISK team.
For additional updates and commentary, please visit http://securityblog.verizonbusiness.com.

AuThoRS: TABle oF ConTenTS
Wade H. Baker executive Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Alex Hutton Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
C. David Hylender State of Cybercrime, 2009. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Christopher Novak Results and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Christopher Porter Demographics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Bryan Sartin Sources of Data Breaches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Peter Tippett, M.D., Ph.D. Breach Size by Source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
J. Andrew Valentine external Breach Sources. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Internal Breach Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
partner Breach Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
ConTRIBuToRS: Threat and Attack Categories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
hacking and Intrusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Thijs Bosschert Malware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
Eric Brohm Misuse and Abuse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Calvin Chang Deceit and Social Attacks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Ron Dormido physical Attacks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
K. Eric Gentry errors and omissions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Mark Goudie Attack Difficulty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Ricky Ho Attack Targeting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Stan S. Kang Compromised Assets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Wayne Lee Compromised Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
Jelle Niemantsverdriet unknown unknowns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

David Ostertag Time Span of Breach events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

Michael Rosen pre-Attack Research. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
point of entry to Compromise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Enrico Telemaque
Compromise to Discovery. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
Matthijs Van Der Wel
Discovery to Containment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Ben Van Erck
Discovery and Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Members of the RISK Team
Discovery Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
ICSA Labs
utilization of Detective Controls. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
Anti-Forensics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
payment Card Industry Data Security Standard . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
SpeCIAl ThAnKS To:
Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Janet Brumfield About the Verizon Business Investigative Response Team . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Carl Grygiel
Hunter Montgomery

1

2009 Data Breach Investigations Report
A study conducted by the Verizon Business RISK team
Executive Summary
2008 will likely be remembered as a tumultuous year for corporations and consumers alike. Fear, uncertainty, and doubt
seized global financial markets; corporate giants toppled with alarming regularity; and many who previously lived in
abundance found providing for just the essentials to be difficult. Among the headlines of economic woes came reports of
some of the largest data breaches in history. These events served as a reminder that, in addition to our markets, the safety
and security of our information could not be assumed either.

The 2009 Data Breach Investigations Report (DBIR) covers this chaotic period in history from the viewpoint of our forensic
investigators. The 90 confirmed breaches within our 2008 caseload encompass an astounding 285 million compromised
records. These records have a compelling story to tell, and the pages of this report are dedicated to relaying it. As with last
year, our goal is that the data and analysis presented in this report prove helpful to the planning and security efforts of our
readers. Below are a few highlights from the report:

Who is behind data breaches?

74% resulted from external sources (+1%). Closely resembling the stats from our 2008 report, most
data breaches continue to originate from external sources.

20% were caused by insiders (+2%). Though still a third of our sample, breaches linked to
business partners fell for the first time in years. The median
size of breaches caused by insiders is still the highest but the
32% implicated business partners (-7%). predominance of total records lost was attributed to
outsiders. 91 percent of all compromised records were
39% involved multiple parties (+9%). linked to organized criminal groups.

How do breaches occur?

In the more successful breaches, the attacker exploited some
mistake committed by the victim, hacked into the network,
67% were aided by significant errors (<>).
and installed malware on a system to collect data. 98 percent
of all records breached included at least one of these 64% resulted from hacking (+5%).
38% utilized malware (+7%).
attributes. unauthorized access via default credentials (usually
third-party remote access) and SQl injection (against web
applications) were the top types of hacking. The percentage
of customized malware used in these attacks more than
doubled in 2008. privilege misuse was fairly common, but not
22% involved privilege misuse (+7%).
many breaches from physical attacks were observed in 2008.
9% occurred via physical attacks (+7%).

2

What commonalities exist?

69% were discovered by a third party (-6%). only 17 percent of attacks were designated to be highly
difficult, yet they accounted for 95 percent of the total records
81% of victims were not payment Card Industry
(pCI) compliant.
breached. So, while hackers prefer soft targets, they do seem to
know where best to apply the pressure when motivated. Most

83% of attacks were not highly difficult (<>).
of these incidents do not require difficult or expensive
preventive controls; mistakes and oversight hinder security
efforts more than a lack of resources. 81 percent of
87% were considered avoidable through simple or
intermediate controls (<>).
organizations subject to pCI DSS had not been found
compliant prior to the breach. nearly all records in 2008 were
compromised from online assets. As with last year’s report, the
99.9% of records were compromised from servers
and applications.
majority of breaches are discovered by a third party.

Where should mitigation efforts be focused?

Some will recognize three of these five recommendations as
carryovers from our previous report. This is intentional. We
simply could not convince ourselves to remove them just to
avoid reiteration. In fact, a fresh look and further
consideration is warranted.
The best defense against data breaches is, in theory, quite
simple—don’t retain data. Since that is not realistic for many
organizations, the next best thing is to retain only what is ensure essential controls are met.
required for business or legal reasons, to know where it lives
and flows, and to protect it diligently.
The majority of breaches still occur because basic controls
were not in place or because those that were present were Find, track, and assess data.
not consistently implemented across the organization. If
obvious weaknesses are left exposed, chances are the attacker
will exploit them. It is much less likely that they will expend
the time and effort if none are readily apparent. Collect and monitor event logs.
As a specific extension of this, we felt it necessary to call out
several tried and true controls based on our 2008 case data. A
very large proportion of attackers gain access to enterprise
networks via default, shared, or stolen credentials.
Furthermore, organizations seem to have little visibility into
Audit user accounts and credentials.
this problem. It’s certainly best to prevent such incidents in
the first place, but a second line of defense is to review
accounts for signs of abuse or anomalies. SQl injection was
also an oft-used means of breaching corporate data last year.
Test and review web applications.
Secure development, code review, application testing, etc. are
all considered beneficial in light of this finding.
Whatever the sophistication and aggressiveness of attacks,
the ability to detect a breach when it occurs is a huge
stumbling block for most organizations. Whether the
deficiency lies in technology or process, the result is the
same—during the last five years, few victims discover their
own breaches. Fewer still discover them in a timely manner.

3

Methodology
The underlying methodology used in this report remains unchanged from the previous year. All results are based on firsthand
evidence collected during data breach investigations conducted by Verizon Business from 2004 to 2008. The 2008 caseload is
the primary analytical focus of the report, but the entire range of data is referenced extensively throughout. Though the
Investigative Response (IR) team works a variety of engagements, only those involving a confirmed breach are included in this
data set. To help ensure reliable and consistent input, all investigators use the same standardized tool to record case data and
other relevant details. This information is then submitted to other members of the RISK team for further validation and analysis.

Beyond this, there are a few notable differences and additions with respect to the 2009 Data Breach Investigations Report.
Whereas the 2008 report reached back across four years of cases in one massive data collection effort, this data set was
assembled periodically throughout the year. Investigators were able to enter information at the close of a case while it was
still fresh in their minds. This shift from historic to ongoing collection allows for more detail on existing data points and opens
the door to new areas of study. We hope these additions enhance the value and utility of this report to the research and
practitioner communities.

Most of the statistics presented in this report refer to the percentage of cases, the percentage of records breached, or simply
the number of cases. The ”percentage of records” statistic is new this year and gives a sometimes different but always insightful
view of the data. Because of the potentially misleading nature of assigning percentages to small samples, the raw number of
cases is used anytime we discuss a subsample within the caseload. For instance, evidence of malware was found in 38
percent of cases, and in the several pages dedicated to these attacks, all figures show integers. Captions and legends should
aid proper interpretation.

Whereas the 2008 report reached back across four years of
cases in one massive data collection effort, this data set was
assembled periodically throughout the year. This shift from
historic to ongoing collection allows for more detail on existing
data points and opens the door to new areas of study.

We would like to reiterate that we make no claim that the findings of this report are representative of all data breaches in all
organizations at all times. These statistics are based solely upon our caseload and any conclusions or inferences we make are
drawn from this sample. Although we believe many of these results to be appropriate for generalization, bias undoubtedly
exists. Even so, there is a wealth of information here and no shortage of valid and clear takeaways. As with any study, readers
will ultimately decide which findings are applicable within their organization.

Finally, it is important to note that Verizon Business is committed to maintaining the privacy and anonymity of Investigative
Response clients. Once the investigator records and submits case metrics, this information is sanitized and the client’s name
is removed from the records. The central repository of case data contains no information that would enable one to ascertain
a client’s identity. Furthermore, the statistics within this report are always presented in aggregate; individual records are never
the focus of analysis.

4

State of Cybercrime, 2009
Before delving into the statistics and analysis presented in our 2009 report, we thought it a good idea to update the “Primer
on Cybercrime” originally presented in the 2008 DBIR. This brief section attempts to put some context around the data and
highlight important aspects of the continuing evolution of cybercrime around the world. One may doubt that the cybercrime
market could change much over a single year, but one need only consider global financial markets in 2008 to realize that
any market system can change and, at times, change swiftly. As the cybercrime market evolves, attackers, targets, and
techniques do as well.

The potential value of engaging in cybercrime would not exist without a market for stolen data. As with any legitimate
market system, the unit value of goods and services fluctuates with supply and demand. Massive exposures of magnetic-
stripe data in recent years (hundreds of millions in our caseload alone) have effectively flooded the information black market,
saturating it with “dumps,” or credit card magnetic stripe sequences sufficient for counterfeit. This market saturation has
driven the price down to a point where magnetic-stripe information is close to worthless. The value associated with selling
stolen credit card data have dropped from between $10 and $16 per record in mid-2007 to less than $0.50 per record today.*

As supply has increased and prices fallen, criminals have had to
overhaul their processes and differentiate their products in order to
As supply has maintain profitability. In 2008, this was accomplished by targeting
points of data concentration or aggregation and acquiring more
increased and prices valuable sets of consumer information. The big money is now in
fallen, criminals have stealing personal identification number (PIN) information together

had to overhaul their with associated credit and debit accounts. Thus, we saw an explosion
of attacks targeting PIN data in the previous year. These PIN-based
processes and attacks hit the consumer much harder than typical signature-based
differentiate their counterfeit attacks. This is because PIN fraud typically leads to cash
being withdrawn directly from the consumer’s account—whether it
products in order to be a checking, savings, or brokerage account. Furthermore, PIN fraud
maintain profitability. typically places a larger share of the burden upon the consumer to
prove that transactions are fraudulent. This makes the recovery of lost
In 2008, this was assets more difficult than with standard credit-fraud charges.
accomplished by The higher value commanded by PIN data has spawned a cycle of
targeting points of innovation in attack methodologies. Criminals have reengineered their

data concentration processes and developed new tools—such as memory-scraping
malware—to steal this valuable commodity. This has led to the
or aggregation and successful execution of complex attack strategies previously thought
acquiring more to be only theoretically possible. As a result, our 2008 caseload is
reflective of these trends and includes more targeted, cutting edge,
valuable sets of complex, and clever cybercrime attacks than seen in previous years..
consumer information.

*Figures based on data collected as part of Verizon Business underground intelligence operations.

5

Results and Analysis
The Verizon Business IR team worked well over 150 forensic engagements in 2008. Of those, 90 were data compromise
investigations in which a breach was confirmed. A number of these investigations were quite extensive and lengthy; a fact
which contributed to the lower-than-average number of cases worked this year. Though fewer, these 90 held their own; the
total number of records breached across our 2008 caseload—more than 285 million—exceeded the combined total from
2004 to 2007.

At the time of this writing, about a third of the breaches investigated by our team last year are publicly disclosed. More,
especially those toward the end of the year, are likely to follow. Others will likely remain unknown to the world as they do
not fall under any legal disclosure requirements.

Roughly 20 percent of 2008 cases involved more than one breach. That is to say, multiple distinct entities or locations were
individually compromised as part of a single case. Amazingly, nearly half of our caseload was comprised of different sets of
interrelated incidents. Quite often the same individual(s) committed the attack. Other times, there was a shared connection
(literally) between the victims and a common third party that experienced a breach. Still others were linked through some
kind of common application, identical attack patterns, and the like.

These 90 cases along with those worked between 2004 and 2007 form the basis of all results and analysis within this report.

Demographics
As with last year’s report, data breaches affected a wide array of organizations in 2008. These are categorized according to the
industry groups presented in Figure 1. Claiming nearly a third of all breaches, retail continues to be the most frequently
affected industry. Food and beverage establishments, second-most common in the 2004 to 2007 data set, dropped in both
proportion (20 percent to 14 percent) and position (now third place) in 2008. The major gainer in 2008 was financial services,
which doubled in terms of caseload percentage to 30 percent.

The increase of data breaches in the financial sector is indicative of recent trends in cybercriminal activity highlighted in the
“State of Cybercrime” section. As will be discussed throughout this report, financial services firms were singled out and fell
victim to some very determined, very sophisticated, and—unfortunately—very successful attacks in 2008. This industry
accounted for 93 percent of the over 285 million records compromised. This finding reflects a few very large breaches

Figure 1. Industries represented by percent of breaches

31% 30% 14% 6% 6% 6% 3% 4%
Retail Financial Food and Manufacturing Business hospitality Technology other
Services Beverage Services

Comments or questions on this section?
Visit http://securityblog.verizonbusiness.com/category/2009dbir/,
and look for the “Demographics” post.

6

investigated by our IR team in the past year. Though few in number,
Figure 2. Industries represented by
they dominate all percentage of records statistics discussed throughout
percent of records
this report.
7% everyone else
Beyond these top three industry groups, a smattering of others filled
out the remaining quarter of cases. Manufacturing and business
services (which includes a few media, marketing, consulting, and legal
firms) and hospitality each accounted for 6 percent of the caseload.
Technology firms, which made up 13 percent of our 2004 to 2007
cases, were comparatively less represented in 2008. We view this
difference to be more reflective of our sample than a broader trend.

The number of investigations handled by our IR team outside the
United States rose to over one-third of our caseload in 2008. In addition 93% Financial Services

to extensive investigations across the United States, many breaches hit
organizations in Canada and Europe while casework demands continued to grow in Brazil, Indonesia, the Philippines, Japan,
and Australia. As attackers continue to pursue soft targets internationally, concern in emerging economies will rise as well,
especially with respect to consumer data.

The distribution of organizational size looks very similar to the previous data set. Per Figure 3, data thieves seem to show no
partiality between larger enterprises and smaller establishments. Though not always the case, criminals typically initiate
attacks based on perceived value of the data and convenience rather than victim characteristics such as size.

One final point of interest deserves mention before concluding this section. A newly added line of inquiry for 2008 found that
13 percent of organizations in our caseload had recently been merged or acquired. It’s difficult to draw a conclusion from this

Figure 3. number of employees by percent of breaches

over 100,000 6%

10,001 to
18%
100,000

1,001 to 10,000 27%

101 to 1,000 17%

11 to 100 26%

1 to 10 7%

7

statistic or assign any significance to it—yet the potential effect of such changes on the likelihood of suffering a breach is
worth considering.

Mergers and acquisitions bring together not only the people and products of once separate organizations but their
technology environments as well. Integration rarely happens overnight or without a hitch. Technology standards are
sometimes set aside for the sake of business expediency. This introduction of variance into the IT operating environment may
serve to increase the risk of compromise. Furthermore, businesses preparing for sale may find reducing operating expenses—
including cutbacks to IT and security spending—a convenient way to help the balance sheet at the time of sale. Finally, new
ownership may alter (by mandate or by culture) the acquired organization’s tolerance for information risk.

All this, of course, is speculation and cannot be proven or disproven (or even tested) without additional information. We
added it to our case metrics with the idea that it might reveal something more substantial over time and we will continue to
record and report it.

Sources of Data Breaches
Similar to cases conducted in the physical realm, one of the primary objectives during a computer forensics investigation is to
identify those responsible for the crime. Because perpetrators often return to the scene, knowing the source of a breach can be
essential to its containment. At a high-level, security incidents originate from one or a combination of the following sources:

External: External threats originate from sources outside the organization. Examples include hackers, organized crime
groups, and government entities, as well as environmental events such as weather and earthquakes. Typically, no trust or
privilege is implied for external entities.

Internal: Internal threat sources are those originating from within the organization. This encompasses human assets—
company executives, employees and interns—as well as other assets such as physical facilities and information systems. Most
insiders are trusted to a certain degree and some, IT administrators in particular, have high levels of access and privilege.

Partner: Partners include any third party sharing a business relationship with the organization. This value chain of partners,
vendors, suppliers, contractors, and customers is known as the extended enterprise. Information exchange is the lifeblood of
the extended enterprise, and, for this reason, some level of trust and privilege is usually implied between business partners.

Results from 600 incidents over five years make a strong
case against the long-abiding and deeply held belief
that insiders are behind most breaches.

and look for the “Sources of Data Breaches” post.

8

Figure 4. Sources of breaches by percent of breaches If evidence reveals that any of these played a significant and
active role in the breach, it is marked as a source. While there is
some room for interpretation in “significant and active,”
74%
investigators do follow a set of consistent guidelines. For instance,
an insider that deliberately steals proprietary information from
their employer is clearly an “internal” breach. We also consider
insiders partially responsible when their actions, though
unintentional, either directly cause or contribute to the breach.
Picking up malware while browsing that is later used by an
32% external attacker to gain unauthorized access is an example of
this. We do not consider it an internal source when an insider’s
20% inaction (i.e., oversight, failure to follow-through on procedures,
decision to not implement certain security measures, etc)
allows or aids a breach. The distribution of breach sources in
external Internal partner 2008 is presented in Figure 4. The results are quite similar to that
Confirmed Suspected of the 2004 to 2007 data set and continue to challenge some of
the prevailing wisdom in the security community with regard to
the origins of data breaches.

Prior to further discussion of these results, it’s worth clarifying two points of potential confusion. First, it is no mistake that the
values in Figure 4 sum to more than 100 percent, as many breaches involve multiple parties. Figure 5 below illustrates the
distribution of breach sources to highlight this fact. Second, we want to be clear that these findings relate specifically to the
occurance (or likelihood) of security breaches leading to data compromise within our caseload—not attacks, not impact, not
general security incidents, and not risk. We observed some rather strong reactions to this finding after last year’s report, and it
was apparent that at least some of the discussion had more to do with terminology than the actual results.

The majority of data breaches continue to originate
from sources outside the victim. In 2008 Verizon Business Figure 5. Single vs. multiple breach sources
by percent of breaches
encountered nearly the same percentage (74 percent)
of confirmed external breaches as our combined 2004
to 2007 caseload. Furthermore, this statistic remains
remarkably consistent over the five-year period of this
study. Based on these results, it seems unwise to 43% 39% 11% 7%
downplay the threat posed by outsiders. only external Multiple Sources only Internal only partner

9

Figure 6. Breach sources over time by percent of breaches

100%

90%

80% external

70%

60%

50%

40% partner

30%

20%
Internal
10%

0%

2004 2005 2006 2007 2008

Insiders, on the other hand, are behind the lowest proportion (20 percent) of breaches in our caseload for four years
running. Figure 5 shows that only about half of these (11 percent of all breaches) were committed by an insider acting
alone. The remainder of the breaches tied to insiders mostly involved employees as unwitting participants in the crime
through errors and policy violations. It is true that these results are based upon our caseload—which is consumer data-
heavy—and may not be reflective of all data breaches. Perhaps insiders are more apt to target other types of data such as
intellectual property. It is also true that many insider crimes may never be detected, though one would think any breach
causing material harm would eventually be noticed. It is also feasible they are more likely handled internally. At any rate,
results from 600 incidents over five years make a strong case against the long-abiding and deeply held belief that insiders
are behind most breaches.

The number of breaches linked to business partners continues to land between external and internal sources but did drop
12 percent in 2008. Interpreting this decline is difficult as it is doubtful that huge strides were made in the effort to reduce
partner-facing risk. It is more likely related to the lower proportion of food and beverage and retail cases within our 2008
caseload. Readers of last year’s supplemental report may remember that those two industries exhibited high percentages of
partner-related breaches (particularly food and beverage at 70 percent or more). In contrast, this year’s results show that
criminals appear to be directly targeting victims that offer a bigger payout. The “end around” maneuver via trusted partner
infrastructure does not seem to be the vector of choice in these attacks. Nevertheless, breaches involving partners are still
quite common and account for over one-third of cases if both confirmed and suspected cases are counted. Any other
difference from past data that cannot be explained due to caseload composition is likely insignificant statistical variation.

10

Breach Size by Source
Figure 7 shows the median* number of records compromised Figure 7. Median number of records compromised
per breach
per event for each threat source. As a reminder, we do not 100,000
assert that the full consequences of a breach are limited to the
number of records exposed; we use this statistic merely as a
measurable indicator of the overall impact.

Insider breaches (individually) continue to be much more
damaging than those caused by other sources though the
difference between them is not to the extent observed across 37,847
our 2004 to 2007 caseload. One of the more interesting changes
27,000
is that outsiders compromised more records per incident than
partners. This shift is attributable to several very large breaches
investigated in 2008 which were perpetrated by outsiders. A
comparison of the median value provided at right (37,847) external Internal partner

with the mean** (5,651,067) gives an appreciation for the
dramatic skew that exists within the data set with respect to
the size of external breaches. This is one of several reasons why we use the median as the preferred measure of central
tendency when analyzing these incidents. Figure 8 provides a striking view of the size and dominant nature of external
breaches last year.

At this point, those familiar with our pseudo risk calculation (likelihood x impact) and its result in the last report may suspect
that it will yield a different outcome this year. That instinct would be correct. Case results from 2008 find that outsiders
represent the greatest risk for data compromise, followed closely by insiders and then partners. This presents a pattern
exactly opposite from what was depicted in our 2004 to 2007 data set. Does this mean that the fundamental nature of
information risk experienced a profound metamorphosis last year? It is doubtful; keep in mind that risk is probabilistic and
best understood over time with multiple measurements. Though few in number, several large breaches were enough to tip
the scales in the direction of outsiders as the dominant source in 2008.

Figure 8. Total records compromised by source

266,788,000 15,796,000 1,509,000 1,330,000

only external Multiple Sources only partner only Internal

*The middle value in an ascending set of numbers
**The average of a set of numbers

11

Table 1. pseudo risk calculation

Impact
Source Likelihood (number of records) Risk (pseudo)
external 74% 37,847 28,175

Internal 20% 100,000 20,000

partner 32% 27,000 8,700

External Breach Sources
The true geographic origin of an attack is difficult to pinpoint with certainty. This determination is predicated upon the source
IP address, which is often unreliable for many reasons. Even so, additional validation is gained through common elements
between cases, correlative fraud patterns, information provided by other Verizon Business departments, and collaboration
with law enforcement agencies. The geographic distribution of external data breach sources is shown in Figure 9.

Though in slightly different order, Eastern Europe, East Asia, and North America remain at the top of the list in 2008. In fact,
these regions are even more dominant, accounting for 82 percent of all external attacks. By comparison, 59 percent of breaches
between 2004 and 2007 originated from these regions. Eastern Asia (up 15 percent) and Eastern Europe (up 9 percent) are
most responsible for the change.

Though it’s tempting to pander to hype
Figure 9. location of attacking Ip(s) by number of breaches
surrounding state-sponsored attacks from
Asia, we find no evidence to support the
position that governments are a significant
east europe 22
agent of cybercrime. We do have a great deal
of evidence that malicious activity from
east Asia 18
Eastern Europe is the work of organized
crime. This is further seen from Figure 10
north America 15
which categorizes external entities into
familiar types rather than by region.
South America 6
That nearly two-thirds are “not traced to a
specific entity other than IP” is the result of South/Southeast Asia 3
several factors. Sometimes we are unable to
do so. Other times the victim decides it is West/South europe 3

not worth the additional time and expense.
In most cases, the immediate need with Middle east 1

respect to the IP address is in containing
the breach rather than rooting out the

12

entities responsible. In those instances when an attempt is
made to trace the IP to a specific entity, we work with law Figure 10. Categories of external breach sources
by number of breaches
enforcement personnel. As seen in the chart, the trail often
leads to members of known organized crime outfits. What
43 19 6
is not evident from Figure 10 is the astounding statistic
that 91 percent of all compromised records in 2008 was
attributed to organized criminal activity. On the brighter
side, we are happy to report that these efforts with law not Traced to Known Known
Specific entity organized Individuals
enforcement led to arrests in at least 15 cases (and other Than Ip Crime
counting) in 2008.

Internal Breach Sources
Several broad classifications of insiders are presented in Figure 11 along with the percentage of incidents attributed to each.
2008 results are similar to the 2004 to 2007 data set. End-users and IT administrators continue to be the culprits behind most
breaches. This finding for IT administrators is not surprising; higher privileges afford greater opportunity and temptation for
abuse. At the same time, the results for incidents perpetrated by end-users serve to remind us that internal breaches are not
solely dependent on privileges or administrative credentials. Though our metrics do include options for part-time and
temporary workers, our caseload included none.

Of all insider cases in 2008, investigators determined about two-thirds were the result of deliberate action and the rest were
unintentional. While it’s tempting to infer that administrators acted more deliberately and maliciously than end-users and
other employees, the evidence does not support this conclusion. The ratio was roughly equal between them. It is worth
noting that both cases involving senior management were the result of deliberate action which was taken after the person
was terminated. We also noticed several other breaches in the caseload were perpetrated by recently terminated employees.
The majority was administrators, but a few cases involved end-users as well. With respect to breaches caused by recently
terminated employees, the following two scenarios were observed:

y Employee was terminated and his/her account was not disabled in a timely manner.

y Employee was notified of termination but was allowed to “finish the day” unmonitored and with normal access/privileges.

This obviously speaks to the need for termination plans that are timely and encompass all areas of access (decommissioning
accounts, disabling privileges, escorting terminated employees, etc.).

Figure 11. Categories of internal breach sources by number of breaches

9 8 2 0 2

no Specific
end-user IT Admin. / Senior Mgt. Agent / Spy person
Super-user Identified

13

Partner Breach Sources
The majority of breaches involving a business partner was the result of third-party information assets and connections being
compromised and used to attack the victim’s systems. This statistic increased substantially in 2008. (For the sake of reference, it
was slightly over half of 2004 to 2007 cases.) In the large majority of cases, it was the lax security practices of the third party that
allowed the attack. It should not come as a surprise that organizations
Figure 12. Categories of partner breach frequently lack measures to provide visibility and accountability for
sources by number of breaches partner-facing systems.

Figure 12 also reminds us that not all data breaches within the
25 6 1
extended enterprise are unintentional. Rising to a slightly higher
proportion last year, six instances of deliberate malicious action by
third-party remote administrators were observed. One of these
no individuals had been recently terminated.
partner Asset or Remote IT Specific person
Connection Admin. Identified After last year’s report, we had many inquiries regarding the nature of
the relationship between the victim (client) and the partner. We
attempted to capture this information during 2008 investigations. This
new information found that most of these breaches dealt with a partner who administered victim-side assets. For retail and
food and beverage organizations, this was almost always a vendor supporting a point-of-sale (POS) system. We also noted
several instances where the partner had user-level access to the victim’s systems or regularly exchanged data with the victim.
Only one case involved a partner physically handling or transporting victim assets. Interestingly, our caseload included zero
instances where the partner hosted the victim systems.

In the large majority of cases, it was the lax security practices of
the third party that allowed the attack. It should not come as a
surprise that organizations frequently lack measures to provide
visibility and accountability for partner-facing systems.
Threat and Attack Categories
Anyone responsible for safeguarding corporate information assets knows there are countless ways in which sensitive
information will find its way into the wrong hands. Though sometimes one-dimensional, data breaches are more often the
result of a series of intertwined and orchestrated events. Examining the frequencies and trends surrounding these scenarios
is essential to protection efforts and is the purpose of this section.

and look for the “Threat and Attack Categories” post.

14

Though very specific attack details are noted during an investigation, all possibilities fit somewhere within the seven top-
level threat categories listed in the figure below. Figure 13 records the prevalence of each as causing or contributing to data
breaches investigated by Verizon Business in 2008 (black bars). Since most incidents involve events spanning several
categories, the percentages sum to well over 100 percent. Also depicted is the percentage of total compromised records
ascribed to each category (red bars).

The results for 2008 cases look very similar to those of the 2004 to 2007 data set (see Figure 14 below for a time series chart
of these results). The Deceit and Physical categories switched places, but all others remained in order. Furthermore, Hacking
and Malware continue to dominate the caseload. Error is seldom the proximate cause of a breach, but it is very often a factor
contributing to or enabling a successful attack. From Figure 13, one can deduce the stereotypical breach scenario: the
attacker takes advantage of some mistake committed by the victim, hacks into the network, and installs malware on a system
to collect data. As evidenced by the red bars in Figure 13, this is especially true for large breaches. The following sections
provide a more in-depth examination of each threat category.

Figure 13. Threat categories by percent of breaches (black) and records (red)

64%
hacking
94%

38%
Malware
90%

22%
Misuse
2%

12%
Deceit
6%

9%
physical
2%

1% 67%
error
0%

0%
environmental % of Cases % of Records
0%

15

Figure 14. Threat categories over time by percent of breaches

100%

90%
hacking
80%
Malware
70%
Misuse
60%

50% Deceit

40% physical
30%
error (Cause)
20%
environmental
10%

0% error (Factor)

2004 2005 2006 2007 2008

Hacking and Intrusion
In terms of malicious action against information systems, hacking is the leading cause of data breaches for the fifth year
running. Since hacking is less subject to the constraints that limit other attack methods (i.e., physical proximity, human
interactions, special privileges), this is not unexpected. Additionally, many tools are available to help automate and accelerate
the attack process, which keeps the cost of attack relatively low for the criminal. Those familiar with attack classification
methodologies will know that the library of hacking and intrusion techniques is quite extensive. For 2008, we expanded our
IR case metrics to provide more detail around this prevalent and potent threat category. Figure 15 reveals the types of
hacking observed by Verizon Business during breach investigations in the last year.

From the chart, it is evident that many intrusions exploit the basic (mis)management of identity. Unauthorized access via
default, shared, or stolen credentials constituted more than a third of the entire Hacking category and over half of all
compromised records. It is particularly disconcerting that so many large breaches stem from the use of default and/or shared
credentials, given the relative ease with which these attacks could be prevented. Readers may wonder why default and
shared credentials are lumped together, as these categories seem to represent two different problems. The answer is that
these issues were frequently found in tandem. We investigated an entire series of cases in which multiple organizations
within the same industry all suffered breaches within a very short timeframe. It didn’t take long to figure out that each used
the same third-party vendor to remotely manage their systems. Unfortunately, that vendor neglected to change the default
username and password—and used the same credentials across multiple clients.

16

Similarly disturbing are those breaches (and the high percentage of compromised records) traced to poor access control lists
(ACLs). In more than a few cases, ACLs proved to be somewhat of a misnomer, leaving a wide-open door for the assailant to
walk through unchallenged. Criminals will usually take the path of least resistance, and unfettered access fits that description
quite well.

When hackers are required to work to gain access, SQL injection appears to be the uncontested technique of choice. In 2008,
this type of attack ranked second in prevalence (utilized in 16 breaches) and first in the amount of records compromised (79
percent of the aggregate 285 million). At its most basic level, SQL injection attacks exploit a failure to properly validate user
input. This seems especially common with custom-developed applications and web front-ends. In the absence of third-party
OS and platform-specific vulnerabilities, criminals are aware of and exploiting weaknesses in application development
processes. SQL injection has been a part of the security industry consciousness for years now, and some may wonder at its
continued prevalence. Fixing vulnerable applications, however, can be challenging, costly, and time consuming, all of which
contribute to a rather large and persistent attack surface. On top of this, SQL injection attacks are growing notably more
sophisticated, especially for data compromise scenarios. It is often used to gain deeper access into systems and plant
malicious software. Also noteworthy relative to hacking techniques is the infrequency with which more commonly known
hacking techniques, such as buffer overflows, exploitation of session variables, and privilege escalation, appear in our data set.

Figure 15. Types of hacking by number of breaches (black) and percent of records (red)

unauthorized Access via Default or Shared
17 / 53 %
Credentials

SQl Injection 16 / 79%

Improperly Constrained or Misconfigured
9 / 66%
ACls
unauthorized Access via Stolen
7 / 0.1%
Credentials

Authentication Bypass 5 / 0.1%

Brute-Force 4 / 7%

privilege escalation 4 / 0%

exploitation of Session Variables 3 / 0%

Buffer overflow 3 / 0%

Cross-Site Scripting 1 / 0%

17

Vulnerability exploits

2008 continued a downward trend in attacks that exploit patchable vulnerabilities versus those that exploit configuration
weaknesses or functionality. Only six confirmed breaches resulted from an attack exploiting a patchable vulnerability. The
word “patchable” here is chosen carefully since we find that vulnerability does not have the same meaning for everyone
within the security community. While programming errors and misconfigurations are vulnerabilities in the broader sense,
lousy code can’t always be fixed through patching and the careless administration patch has yet to be released. Furthermore,
many custom-developed or proprietary applications simply do not have
routine patch creation or deployment schedules. Table 2. patch availability at time of breach
For the six exploited vulnerabilities that had existing patches available,
Table 2 shows how long the patch had been public at the time of the less than 1 month 0
breach. The story is similar to that of the previous report; the interim
1 to 3 months 0
between a patch’s release and active exploits leading to data compromise
is usually on the order of years. Vulnerabilities are certainly a problem 3 to 6 months 0
contributing to data breaches, but patching faster is not the solution.
6 to 12 months 1
This year’s findings continue to support the idea that a patch deployment
strategy focusing on coverage and consistency is far more effective at More than 1 year 5
preventing data breaches than “fire drills” attempting to patch particular
systems as quickly as possible.

Attack Vector

Looking deeper into hacking activity, it is apparent that the bulk of attacks continues to target applications and services
rather than the operating systems or platforms on which they run. Of these, remote access services and web applications
were the vector through which the attacker gained access to corporate systems in the vast majority of cases. While network
devices do sometimes serve as the avenue of attack, it was considerably less often in 2008.

In approximately four of 10 hacking-related breaches, an attacker gained unauthorized access to the victim via one of the
many types of remote access and management software. Rather than for internal usage, most of these connections were
provisioned to third parties in order to remotely administer systems. As discussed extensively in this and previous reports, the
ultimate attacker is not typically the third party (although that certainly happens). More often, an external entity compromises
the partner and then uses trusted connections to access the victim. From the victim’s perspective the attacker appears to be
an authorized third party, making this scenario particularly problematic. This is especially so when trusted access is coupled
with default credentials.

Although web application attacks are one fewer in number than those against remote access services, they are responsible
for a much larger number (79 percent) of breached records. Based on earlier discussion, it is not difficult to surmise that SQL
injection is the predominate type of attack against this vector. Interestingly, the reason SQL injection is so successful is related
to the scenario described above pertaining to remote access and management software. To function properly, a trust
relationship must exist between web applications and back-end databases. In this sense, a request from the application is

18

Figure 16. Attack pathways by number of breaches (black) and percent of records (red)

Remote Access & Mgt. 22 / 27%

Web Application 21 / 79%

other Server or Application 7 / 7%

network Devices 6 / 11%

end -user Systems 1 / 26%

similar to a request from a privileged administrator. The database obediently yields information requested by the application
and cares not whether the command is valid or the result of an external attacker passing illegitimate strings. This is one of the
primary reasons why encrypting databases is of limited effectiveness in preventing attack scenarios that do not involve physical
theft or control of the system.

A much smaller percentage of hacks targeted routers, switches, and other network devices. This includes wireless networks,
which continue to be a rare attack vector for data breaches. 2008 saw only a single instance in which a wireless network was
exploited across our entire caseload (0.01 percent of all breached records in 2008). An equal number was observed in 2007.
By comparison, 13 percent of cases investigated between 2004 and 2006 involved wireless networks. Incidentally, all were
legitimate corporate WLANs rather than rogue devices deployed without proper authorization. There are many reasons for this
decline, but better out-of-the-box security, wider use of

A much smaller percentage of hacks encryption, and the necessity for proximity in order to
conduct an attack are just a few. Web-based applications
targeted routers, switches, and other and remote access tools are, by their very operational
network devices. 2008 saw only a single nature, much more visible and accessible to external
entities seeking a way into corporate networks.
instance in which a wireless network was
exploited across our entire caseload.

19

Malware
ICSA Labs, an independent division of Verizon Business, provides credible third-party product testing and certification within
the information security industry. When a member of the IR team discovers malicious software, or malware, during an
investigation, it is sent to ICSA Labs for analysis. Investigators can then use this analysis to better help the customer with
containment, removal, and recovery. The information that follows is based on this collaborative research.

During 2008, malware was involved in over one-third of the cases investigated and contributed to nine out of 10 of all records
breached. In years past, malware was generally delivered in the form of self-replicating email viruses and network worms. The
primary goal was rapid and widespread propagation, typically resulting in availability losses and extensive clean-up. In the
last five years, these goals have shifted. Malware is now an essential component to nearly all large-scale data breach scenarios.
Hacking gets the criminal in the door, but malware gets him the data. Naturally, the criminal will then want to minimize the
chance of detection in order to maximize the amount of data stolen. For this reason, malware becomes ever more directed,
innovative, and stealthy.

By a wide margin, the most common malware delivery method was the scenario in which an attacker compromised a system
and then installed malware on it remotely. Perhaps more importantly, this delivery method accounts for 89 percent of the
records breached in 2008. Seven infections occurred via websites, and, of these, four were “drive-by” downloads requiring no
user interaction. The other three were explicitly downloaded and installed by employees. Only four cases in 2008 involved
malware that exploited a patchable vulnerability. In all of these instances, the necessary patches to prevent infection were
older than one year.

An important corollary to the infection vector is what the malware does once it is placed within the victim environment. In
our previous report, we asserted that most malware captures and stores data locally, captures and sends data to a remote
entity, or enables remote access to or control of the infected system. We also claimed that the ratio between these three
functions was roughly equal. As seen in Figure 18, malware observed in 2008 exhibits a similar result.

Figure 17. Malware infection vector by number of breaches

18 7 1 1 1 6

Installed By Web physical network
email unconfirmed
Attacker Browsing Vector propagation

20

Figure 18. Malware functionality by number of breaches

Keylogger or Spyware 17

Backdoor or Command Shell 16

Capture and Store Data 13

Attacks other Systems 2

Disables Security Controls 2

other 2

The most prevalent of these functions were keyboard loggers or spyware. Typically, these are utilized to capture authentication
credentials which are almost always sent to a remote attacker rather than stored locally for later retrieval. This is probably due
to the small packet size that has a better chance of undetected egress. Criminals often use these credentials for subsequent
and/or expanded attacks against corporate systems.

Responsible for 82 percent of total breached records in 2008, the most effective type of malware in terms of harvesting
massive amounts of sensitive data is the “capture and store” variety. Attackers typically prefer this functionality for breaching
payment card data and personally identifiable information (PII) since frequent exports of huge files containing millions of
records is not the stealthiest of tactics. Of course, storing the payload on the victim’s systems introduces its own challenges—
namely, how to retrieve it. To solve this problem, the attacker will typically open up a backdoor in order to return to the
system undetected over the (ordinarily) months that pass before the jig is up. Among our 2008 cases, investigators found
backdoor or command shell tools in every instance where malware was capturing data to a local file.

The evolving use and functionality of malware in modern data compromise scenarios stems from the cybercrime market
pressures described earlier in this report, but is also a direct response to the widespread adoption of various compliance
standards and requirements. Organizations are implementing prescribed control measures in the manner and extent
required to achieve and maintain compliance. Overall, this is a good thing for data protection within these organizations as
common points of failure are (slowly but surely) being addressed. For instance, organizations are beginning to store less
sensitive data as a part of normal business operations and encrypt what data they do retain. Less unencrypted information
is flowing over public and private networks. Unfortunately, the criminals are not sitting around sulking about lost opportunities
and dead-end business models; they are adapting.

21

As organizations move to meet regulatory requirements, we have observed a manifest increase in attacks designed to
circumvent certain controls implemented as part of that compliance process. Newer, more elaborate varieties of malware
utilities bypass existing data controls and encryption, effectively creating vulnerable data stores that can later be retrieved
from the victim environment. Examples of this include
the usage of memory scrapers, sophisticated packet
capture utilities, and malware that can identify and Unfortunately, the criminals are not
collect specific data sequences within unallocated disk sitting around sulking about lost
space and from the pagefile.
opportunities and dead-end business
Traditionally the term “stored data” has referred to
nontransient items (i.e., in a log file or within a database on
models; they are adapting.
a hard drive, CD, or backup tape). However, the transient
storage of information within a system’s RAM is not typically discussed. Most application vendors do not encrypt data in
memory and for years have considered RAM to be safe. With the advent of malware capable of parsing a system’s RAM for
sensitive information in real-time, however, this has become a soft-spot in the data security armor.

This expanded functionality, of course, doesn’t happen magically. It requires authoring new malicious programs or significant
modification of existing ones. This, in turn, requires considerable amounts of time, money, and expertise—not an investment
many are willing or prepared to make. However, backed by the plenteous resources of organized crime and driven by the
prospect of large hauls of valuable data, it is getting done. This trend was apparent in 2008, during which the percentage of
customized malware more than doubled to 59 percent of all samples encountered.

Figure 19. Malware customization by percent of breaches involving malware

59%
60%

50%

40%

29%
28%
30%

20% 24%
21%

10%

0%
2004 2005 2006 2007 2008

22

Among these cases, the amount and type of Figure 20. Malware customization by number of breaches
customization varied. A very general representation
of this is given in Figure 20. Some attackers simply
repacked existing malware so as to make its signature
undetectable by antivirus software (AV) scanners. 8
Custom-Created
Others leveraged existing malicious code, but modified Code 14
No
it for additional functionality or tailored it to the victim’s Customization
environment. Most common in 2008, however, was
malware that had (apparently) been created for the
attack(s) entirely from scratch. In a rather sobering
5
statistic, 85 percent of the 285 million records breached
Code
in the year were harvested by custom-created Modification
5
malware. It is possible that the code preexisted yet
Repacked to
went unrecognized by the experts and tools at ICSA Avoid AV

Labs, but this matters little to the overall point.

More to the point is that, besides being more capable and better adapted, most malware used for the purpose of
compromising data is not detectable by modern AV. Unfortunately, many organizations rely on AV as the primary means of
malware prevention and detection. AV is certainly a foundational control, but the continuing evolution of malware leaves
security programs built solely upon AV for combating malware unstable at best.

Misuse and Abuse
Misuse refers to the use of organizational resources and/or privileges for any other purpose than for that which they were
originally intended. For this reason, the category is particular to insiders and partners, as they are trusted by the organization
to some degree.

Overall, 22 percent of breaches were caused by some form of misuse (see Figure 13). We find the fact that insider and partner
misuse only accounted for 2 percent of all records compromised in 2008 to be counterintuitive since breaches involving the
abuse of system privileges are usually quite damaging. 2008 was a rather unusual year, however, and this finding squares

Newer, more elaborate varieties of malware utilities
bypass existing data controls and encryption, effectively
creating vulnerable data stores that can later be
retrieved from the victim environment.

23

perfectly with the results discussed earlier with respect to internal Table 3. Types of misuse by number of breaches
breaches. Furthermore, such attacks tend to be more narrowly
focused (one only steals what one intends to use) and target Abuse of system access/privileges 15
data types that are not suited to the number of records
Violation of other security policies 6
measurement. One “record” of IP can be quite damaging but
drowned out by millions of payment card numbers. Violation of pC/email/web use policies 5

Table 3 shows the types of misuse observed among these cases.
embezzlement 2
Not surprisingly, the abuse of system access and privileges was
common. Most of these are committed by insiders with
administrative privileges and are deliberate and malicious in
Table 4. Types of assets misused by number of breaches
nature. Nonmalicious policy violations, which are undoubtedly
far more common for the whole of security incidents than these
Database server 6
breach-specific numbers reflect, do contribute to data loss
events but to a lesser extent. Such activity is very often a vector Application server 5
for the introduction of malware into the organization. Also
laptop 5
observed were two instances of embezzlement.

The data in Table 4 shows that when employees engage in File server 3

misuse, they tend to target larger data repositories. The majority
public kiosk system 2
of server-related breaches resulted from privilege abuse, while
incidents involving workstations and laptops were associated poS system 2
with policy violations. It is interesting to note that whereas other
Workstation 2
studies have found portable media to be the leading cause of
data breaches, we observed only a single instance in which such portable media 1
devices were used. Furthermore, in this particular case, the
success of the breach did not hinge on its use; the USB media
was merely a convenient method of moving data (which in our Table 5. Vectors of deceit and social attacks by number
determination would have occurred anyway using other means of breaches

if it weren’t available).
email 5
Deceit and Social Attacks
In-person 4
This category encompasses the use of deception or misrepre-
sentation to exploit people, security measures, procedures, or Web/Internet 3

anything else that furthers the goal of data compromise. These
phone 1
actions can be conducted through both technical and non-
technical means. Common examples of deceit include social Media or paper 1
engineering and phishing scams, both of which we observed in
Instant messaging 0
the 2008 data set. Deceit was apparent in only 12 percent of our
cases and these actions resulted in the compromise of 6 percent
of all records.

24

Table 6. Targets of deceit and social attacks by number The majority of deceitful activity was carried out through email
of breaches
and/or the Internet. However, criminals have not become entirely
impersonal; several instances of social engineering involved
end-user 8
face-to-face and over-the-phone interaction. Web and email
partners or customers 2 served as the medium of phishing and other scam-type attacks.

End-users proved to be the primary target of attacks employing
human resources 1
deceit which, along with other findings in this report, call for more
IT Admin. or Super-user 0 effective security awareness programs at the end-user level.

helpdesk 0
Physical Attacks
Senior management 0 Data breaches resulting from physical attacks continue to rank
low on our list, a finding that some will perceive to be out of sync
with public sources of breach statistics. There are several
Table 7. Types of physical attacks by number of breaches justifiable reasons for this, all of which relate to our caseload. First,
many physical incidents such as lost or stolen laptops never
Theft of asset 5 result in data compromise yet certain regulations require that the
incident be publicly disclosed. Our data set, on the other hand, is
System access (via keyboard) 2
comprised only of incidents in which an actual breach occurred.
equipment or system tampering 2 This has a very dramatic effect on results. Second, the nature of
physical events often precludes the need for third-party
loss or misplacement of asset 0 investigation since there is less forensic evidence available for
Wiretapping 0 examination.

Only eight breaches in 2008 involved a physical attack. These are
observation or shoulder surfing 0
broken down in Table 7 (the values sum to nine because one
Assault or threat of harm 0 case involved two distinct actions). Instances of theft involved
workstations, documents, and backup tapes (no laptops) while
system access was associated with online data repositories. Both
cases of tampering dealt with PIN entry devices attached to POS
systems in retail stores.

In addition to the type of physical attack, investigators noted
where these breaches took place. Here again, the values exceed
the number of breaches. This was due to one fairly elaborate
scheme that unfolded over several venues. The main observation
is that the ratio of external locations to internal locations is
roughly equal.

25

Table 8. locations of physical attacks by number of breaches

Victim location: indoor, publicly accessible area (i.e., lobby, loading dock) 3

external location: store, restaurant, etc. 2

external location: employee home or car 2

external location: airport, train, subway, etc. 1

external location: business partner facilities 1

Victim location: indoor workspace (i.e., cubes, offices) 1

Victim location: indoor high security area (i.e., server room, R&D labs) 1

Victim location: outdoor, corporate grounds (i.e., parking lot) 0

Errors and Omissions
In its broadest sense, error is a contributing factor in nearly all data breaches. Poor decisions, misconfigurations, omissions,
noncompliance, process breakdowns, and the like undoubtedly occur somewhere in the chain of events leading to the
incident. Because error is so incredibly prevalent, only those errors which directly caused or significantly contributed to the
compromise are considered by investigators for inclusion in this report. This distinction between error as a “direct cause” (or
threat) vs. error as a “contributing factor” (failure in protection) is the reason behind the solid (cause) and shaded (factor)
portions of the error bar in Figure 13. Intuitively, error is far less often the direct cause of a data breaches than it is one of
several factors leading to its occurrence. In addition to causal and contributory, we further divide error into the types shown
in Figure 21, which provides a relative distribution of each across 2008 cases in which an error was noted.

Figure 21. Types of error by number of breaches

Misconfiguration 31

omission 31

programming error 16

user error 3

Technical Failure 1

26

Our use of omission and misconfiguration likely warrants some clarification. If an organization enacts a policy or procedure and
then fails to follow through, we consider it an omission. In this sense, omission is not just another way to say the organization
lacks adequate security measures (which is not considered an error in our approach) and is distinct from misconfiguration,
which is a more active form of error during deployment and routine administration of systems. Whereas omissions dominated
the 2004 to 2007 results, the scales are more balanced in 2008. We’d like to attribute this to dramatically improved assurance
mechanisms over the last year, but—unfortunately—it is due to a methodological change on our part. Because the omissions
category was so overpowering, we crafted a more strict definition and also added a new category for programming errors.

Poor decisions, misconfigurations, omissions,
noncompliance, process breakdowns, and the
like undoubtedly occur somewhere in the chain
of events leading to the incident.
As Figure 21 illustrates, misconfiguration was the leading category of error contributing to data compromise, edging out
omissions by a slight margin. 2008 saw quite a few instances of misconfigured or improperly constrained ACLs on perimeter
devices. The use of default passwords despite policies forbidding them was a common example of omission. Clearly, procedures
designed to identify and rectify these common mistakes would go a long way toward shoring up security efforts in most
organizations. 21 percent of recorded errors stemmed from poor coding practices. Not surprisingly, errors of this variety are
strongly related to the occurrence of SQL injection and other similar attacks. Writing secure code is not easy, but training and
tools do exist, and—based on these results—developers would do well to avail themselves of both.

Attack Difficulty
Figure 22. Attack difficulty by percent of breaches
The relative difficulty of attacks leading to data compromise is
not only an excellent indicator of the current threat environment,
10%
but also the state of modern security programs. During each case, None
our investigators assess the details of the attack and classify it
according to the following difficulty levels: 17%
High* 42%
y none: No special skills or resources required. The average user Low

could have done it.
31%
y low: Basic methods, no customization, and/or low resources Moderate
required. Automated tools and script kiddies.

*Highly difficult attacks accounted for 95% of all compromised records.

and look for the “Attack Difficulty” post.

27

y Moderate: Skilled techniques, some customization, and/or significant resources required.

y high: Advanced skills, significant customization, and/or extensive resources required.

Although the rating admittedly involves a dose of subjectivity, we believe it to be a worthwhile metric.

One of the telltale findings from the previous DBIR was that more than half of breaches were caused by rather unsophisticated
attacks. Within a few percentage points, our 2008 cases reveal a very similar statistic. The proportion of breaches requiring no
special skills or resources (None) rose while low difficulty attacks fell to offset that gain. Attacks of moderate difficulty increased
a few points and the percentage of highly difficult attacks remained the same. These subtle shifts are likely inconsequential
and give little reason to believe that attack paradigms changed in
2008. On the whole, it would appear that criminals are still not
Figure 23. Attack difficulty by percent of records
required to work very hard to breach corporate information systems
and 2008 was just a year like any other. Appearances, however, can
5% none/low/Mod
be deceiving.

While it may be true that the majority of breaches are not the result of
highly skillful attacks, an alternate view of the data suggests that the
really high-value targets require extensive effort. As Figure 23 plainly
and powerfully demonstrates, these relatively few highly difficult
attacks compromised 95 percent of the 285 million records across
our caseload—a truly stunning statistic and one that is part of a
larger story.

As discussed in the section detailing hacking activity, the techniques
95% high
used by criminals to infiltrate corporate systems remain relatively
low in sophistication. Exploiting default credentials and misconfigured
ACLs does not require a great deal of skill but can be as effective a
means of entry as more elaborate attacks. Increasingly, the complexity of an attack lies in the challenge of capturing and
compromising the data and perpetuating access to systems. Since malware is typically the means by which this is
accomplished, it has become the difficult component of modern attacks. Difficult attacks, therefore, are not necessarily
difficult to prevent. It may sound trite, but the message here is that attacks are most efficiently and effectively prevented
earlier rather than later.

In the 2008 DBIR, we stated that given enough time, resources and inclination, criminals can breach virtually any single
organization they choose but do not have adequate resources to breach all organizations. Therefore, unless the value of the
information to the criminal is inordinately high, it is not optimal for him to expend his limited resources on a hardened target
while a softer one is available. These statistics make it clear that skilled cybercriminals are foregoing easy pickings in favor of
inordinately high-value targets. Unfortunately, in 2008, they reaped the fruits of their labors.

28

Attack Targeting
Standard convention in the security industry classifies attacks into Figure 24. Targeted vs. opportunistic attacks
by percent of breaches
two broad categories: opportunistic and targeted. Due to significant
grey area in this distinction, we find it useful to separate opportunistic
attacks into two subgroups. The definitions are provided below:

y Random opportunistic: Attacker(s) identified the victim while Fully opportunistic
Targeted* Random
searching randomly or widely for weaknesses (i.e., scanning large 28% 28%
address spaces) and then exploited the weakness.

y Directed opportunistic: Although the victim was specifically
opportunistic
selected, it was because they were known to have a particular (Directed)
44%
weakness that the attacker(s) could exploit.

y Fully Targeted: The victim was first chosen as the target and then
the attacker(s) determined a way to exploit them. *Targeted attacks accounted for 90% of all compromised records.

Given the message in the previous section that criminals seem to be selecting high-value targets and constructing elaborate
attacks designed to breach their defenses, one might infer that targeted attacks rose in 2008. Based on our case evidence,
that would be a valid inference. In fact, targeted attacks are at a five-year high and accounted for 90 percent of the total
records compromised. In a comparison that speaks volumes about the uniqueness of 2008 in this regard, only 14 percent of
the roughly 230 million records exposed among breaches we investigated from 2004 to 2007 were the result of fully targeted
attacks. Financial services organizations are often singled out by criminals due to the large amounts of consumer data they
process, transmit, and store.

As discussed in last year’s report, we encounter many breaches that seem neither truly random nor fully targeted—particularly
in the retail and food and beverage industries. In a very common example, the attacker exploits Software X at Brand A Stores
and later learns that Brand B Stores also runs Software X. An attack is then directed at Brand B Stores but only because of a
known exploitable weakness.

Compared to previous years, fewer breaches in 2008 resulted from purely random searching or scanning for exploitable
weaknesses. One of the fundamental self-assessments every organization should undertake is to determine whether they
are a Target of Choice or Target of Opportunity. If the former, expect and prepare for determined and sophisticated attacks. If
the latter, minimize the opportunities presented so as to become less of a beacon for attack. At the very least, make sure your
beacon shines less brightly than everyone else’s.

One of the fundamental self-assessments every
organization should undertake is to determine whether
they are a Target of Choice or Target of Opportunity.

and look for the “Attack Targeting” post.

29

Compromised Assets
We now turn attention to the types of information assets compromised during these attacks. The focus here is specifically on
the endpoint rather than the pathway of the attack. To allow for comparison with the original report, the same asset groups
are used in Figure 25. A more specific classification is provided in Table 7 for those interested in further details on these assets.

By a large margin and for the fifth year in a row, online data (consisting of various types of servers and applications) is the
most frequently compromised asset. Furthermore, this group accounts for nearly all (99.9 percent) of the 285 million records
breached across our 2008 caseload. As
seen in Figure 25, this percentage
Figure 25. Asset classes by percent of breaches (black) and records (red)
represents a high mark for the five-year
period of this study. Clearly, large and
online Data 94% / 99.9% remotely accessible stores of data remain
the target of cybercriminal activity.
end-user Systems 17% / 0.01%
Looking a bit closer at online data in
offline Data 2% / 0.04% Table 9, POS systems were most fre-
quently compromised but accounted
networks & Devices 0% / 0%
for only a small portion of total records.
Intuitively, these breaches predominantly
afflicted the retail and the food and
beverage industries. Databases rank second in terms of caseload but yielded the majority of breached data. All together,
other types of online data listed in Table 9 factored into a third of breaches, but—of those—only the application servers had
substantial losses of data.

One final point of interest concerning online data is that of virtualization. After the release of last year’s report, there was
some discussion on what our results meant for virtualization. As such, we made a point to note whether assets involved in a
case were virtualized. We encountered five instances among 2008 cases but none were believed to have contributed to the
breach in any way.

Although much angst and security funding is given to offline data, mobile devices, and end-user systems, these assets are
simply not a major point of compromise within the data set available to us for examination. It is indisputable that employees
misuse portable media and laptops go missing, and—based on public breach disclosure lists like DatalossDB.org* and ID
Theft Center**—it is also evident that large numbers of records are reported exposed from related incidents. That such trends
are not reflected in the results above is a by-product of our caseload, this data set, and the general nature of data compromise.
Verizon Business is not often engaged to investigate lost devices and this data set is culled down to only cases in which a
breach was confirmed. Furthermore, information on lost or stolen laptops and media is rarely accessed and abused by
criminals; data-at-risk—though often necessary to report—is not the same as actual data compromise.

*http//datalossdb.org
**http://www.idtheftcenter.org

and look for the “Compromised Assets” post.

30

Figure 26. percent of records breached from online data assets

100%
89% 99.9%

90%

80%
80%
81%
73%
70%

60%

2004 2005 2006 2007 2008

Table 9. Detailed listing of compromised assets by percentage of breaches and records

Asset Asset Group % of Breaches % of Records
poS system online Data 32% 6%

Database server online Data 30% 75%

Application server online Data 12% 19%

Web server online Data 10% 0.004%

File server online Data 8% 0.1%

public kiosk system online Data 2% 0.4%

Authentication / Directory server online Data 2% 0.1%

Backup tapes offline Data 1% 0.04%

Documents offline Data 1% 0.000%

Workstation end-user System 8% 0.01%

laptop end-user System 4% 0.000%

pIn entry Device end-user System 2% 0.004%

31

Compromised Data
2008 was a milestone year in terms of the sheer number of records compromised in breaches investigated by Verizon
Business. More records were breached in 2008 than any other single year, and—astoundingly—more than in the previous
four years combined. In the last five years, our team has investigated cases totaling more than half a billion records.

That 2008 is such an anomaly in this regard is the result of a few very large breaches. The top five breaches account for 93 percent
of total records compromised. The mean number of records per breach was approximately 4.5 million, while the median was
37,847. From these statistics, the data set is obviously skewed right, which can also be seen in the distribution in Figure 28.

As we pointed out last year, one of the more critical components of the investigative process is determining the types of data
compromised as a result of the breach. This determination largely decides the organization’s response strategy. For instance,
certain types of data require public disclosure and/or notification. Some are highly regulated while others require monitoring
for fraudulent activities. In addition to the victim organization, millions of individuals could potentially be affected by the
compromise, and they must be informed as well.

Overall, the results for 2008 are fairly consistent with those of the previous four years. As a percentage of caseload, payment
card breaches remain near the 80 percent mark and far outnumber other data types. They consume 98 percent of all records
compromised in the year. While other types of data are sought by certain groups (i.e., competitors may target IP), the vast
majority of cybercriminals are looking for a quick and easy
payoff. Payment cards certainly fit the bill. As a testimony to
Figure 27. number of records compromised per year in
breaches investigated by Verizon Business this, fraudulent use of stolen card data was confirmed in 83
percent of our cases.
285,423,000
The payment card data percentage above includes not
only the typical data sequences sufficient for card-present
and card-not-present fraud but also PIN information
associated with consumer payment card accounts. The
latter was increasingly targeted in 2008. These attacks
104,321,000 involve the identification and compromise of stores of
75,835,000
magnetic-stripe data, together with PIN, setting the stage
39,042,000 for more damaging forms of identity fraud. This includes, for
11,488,000
example, counterfeit ATM withdrawals and other PIN-based
2004 2005 2006 2007 2008 transactions, often leading to actual cash and related assets
being stolen directly from the consumer. While statistically
not a large percentage of our overall caseload in 2008, attacks

and look for the “Compromised Data” post.

32

Figure 28. Distribution of breach size against PIN information represent individual data theft cases
by number of records
having the largest aggregate exposure in terms of unique
39% records. In other words, PIN-based attacks and many of the
very large compromises from the past year go hand in hand.

PII was the second-most compromised data type, which
only stands to reason as it is also useful in fraudulent activity.
As the name implies, PII includes elements of a person’s
19%
unique identity such as name, identification number, etc. At
30 percent of cases, stolen authentication credentials in
14%
13% 2008 doubled the 2004 to 2007 figure. Whether this is
indicative of some change in criminal strategy remains
8% 8%
unknown (it had been declining each year prior to 2008).
Credentials certainly give attackers the prospect of
increased access for illicit activity and based on the results
< 1K 1K-10K 10K-100K 100K-1M 1M-10M > 10M
from the hacking section earlier in this report, they appear
to be exploiting that advantage.

Though still a smaller share of breaches, theft of intellectual property within our caseload rose to a five-year high in 2008.
Such incidents, while infrequent, can be business-changing events. Account numbers and monetary assets were added to
the list of data types in 2008. Neither was extremely common but multiple instances of each were observed. Account
numbers (often bank accounts) are distinct from payment card numbers. For obvious reasons, these two data types often
appeared in concert; millions of dollars were stolen directly from compromised accounts in 2008.

Figure 29. Compromised data types by percent of breaches (black) and records (red)*

81% / 98%
payment Card Data

personal Information 36% / 1.5%

Authentication Credentials 31% / <0.1%

Account numbers 16 / 0.5%

Intellectual property 13%

Monetary Assets / Funds 11%

Corporate Financial Data 6%

other 11%

*Due to the nature of certain data, not all types listed in Figure 29 are included in percentage of records statistics.

33

Unknown Unknowns
Several years ago, a strong pattern emerged among cases Figure 30. unknown unknowns by percent of breaches
wherein the investigation inevitably led to unknown, 67%

overlooked, or forgotten assets. At some point, we took to
referring to these frequent scenarios as “unknown unknowns,”
53%
and they include any of the following:

y A system unknown to the organization (or business
group affected) 38%

y A system storing data that the organization did not
know existed on that system 24%

y A system that had unknown network connections 17%
15%
or accessibility

y A system that had unknown accounts or privileges
1% ~0%
Figure 30 shows the percentage of 2008 cases in which unknown unknown unknown unknown
each of these contributed to a data compromise. Asset Data Connection privileges
% of Breaches % of Records
Roughly half of all breaches investigated in 2008 revealed at
least one type of unknown. Though substantial, this statistic
is considerably lower than the 90 percent given for the
previous four years. A completely unknown asset factored into only one case, while unknown connections remained near
the 25 percent mark and unknown privileges rose by 7 percent. Far fewer breaches involved data that the victim did not
know existed on the system and this fact accounted for most of the difference between the two data sets. Rather than some
kind of sampling effect, there are several legitimate reasons for this change.

You may remember from the 2008 Data Breach Investigations Supplemental Report that financial services organizations were
less afflicted with unknowns than other industries. Because these organizations represent a higher proportion of our caseload
in 2008, it stands to reason that they would have a downward pull on this statistic. Incidentally, this may also explain the rise in
unknown privileges and drop in unknown connections. Second, based on findings from other sections of this report, it is
evident that we observed a higher-than-normal number of highly targeted and customized attacks this year. The criminals
appear to be going for the “crown jewels,” and the victim typically knows where those reside. Rather than the result of unintended
data migration (though this certainly occurs), more of these breaches stem from a combination of the criminal’s resolve and the
victim’s failure to adequately protect systems they know to contain sensitive data. Still, many did involve unknown data of a
sort—but it was data captured by malware and stored on the system unbeknownst to the organization.

and look for the “Unknown Unknowns” post.

34

Though breaches involving unknown data have seen decline, poor account and privilege management has become more
problematic in recent years both as a percentage of cases and in the number of compromised records. Additionally, Figure 30
reminds us that the consequences of the “unknown unknown” remains high. This suggests that increasing visibility and
reducing variability in the IT operating environment should be a top priority of risk management efforts.

Time Span of Breach Events
The timeline of events leading up to and following a data breach varies greatly depending on a multitude of factors. Some
attacks unfold rapidly while others require lengthy planning and execution. For the purposes of analysis, we separate an
incident into three major phases: point of entry to compromise, compromise to discovery, and discovery to containment.
Information regarding the amount of time spent on pre-attack research is also of interest and recorded when obtainable. The
figures below depict the span of time elapsed during each phase.

Figure 31. Time span of breach events by percent of breaches

Minutes hours Days Weeks Months Years unsure

pre-attack Research

4% 26% 15% 11% 26% 0% 19%

point of entry to Compromise

27% 21% 29% 17% 6% 0% 0%

Compromise to Discovery

0% 8% 16% 25% 49% 1% 0%

Discovery to Containment

0% 6% 37% 42% 15% 0% 0%

and look for the “Time Span of Breach Events” post.

35

Pre-Attack Research
Though solid evidence of pre-attack activity is often elusive, investigators attempt to discern as much as possible about an
attack prior to the point of entry. Doing so often uncovers hints as to how the target was acquired, what reconnaissance
methods were employed, and—most importantly—what signs might have forewarned the victim of an impending attack.

In just under half of our cases in 2008, investigators found at least some indication of pre-attack research. This most often
involved basic system footprinting, scanning, and enumeration, but instances in which the attacker scouted the victim’s
premises were observed as well. The latter were designed to uncover the make and model of POS systems or study traffic
patterns to identify the optimal time for tampering with the PIN Entry Device (PED). In several cases, the assailant was able to
obtain private information (a vendor’s customer lists, for instance) and leveraged it to further customize or to broaden an
attack. Per Figure 31, the time required to conduct these activities varied from minutes to months.

Point of Entry to Compromise
After breaching the perimeter, intruders typically explore the victim’s network and systems until finding their desired plunder.
The length of time necessary to accomplish this largely depends on factors such as the attacker’s prior knowledge, skill set,
familiarity with the environment, the method of attack, and the strength of the victim’s defenses. Overall, 2008 cases follow
a very similar distribution to those in our 2004 to 2007 data set. Though more compromises occurred within minutes, fewer
required hours of effort. Criminals can still access the data in a matter of minutes or hours about half of the time. The
proportion corresponding to days, weeks, months, and years remained nearly identical.

Compromise to Discovery
On the whole, organizations discovered breaches slightly quicker in 2008. However, lest we confuse “quicker” with “quickly,”
this statement needs some additional context. Breaches still go undiscovered and uncontained for weeks or months in 75
percent of cases. It is doubtful that any chief security officer anywhere would call this “quick.” Nevertheless, rather than the
supermajority observed in previous years, 2008 investigations revealed that only a slight majority of breaches took months
or years to discover. On the positive side, we saw comparatively more weeks instead of months and the number detected
within hours increased by 5 percent. We can only hope such trends will continue.

On the whole, organizations discovered breaches slightly
quicker in 2008. However, lest we confuse “quicker” with
“quickly,” this statement needs some additional context.
Breaches still go undiscovered and uncontained for weeks
or months in 75 percent of cases. It is doubtful that any
chief security officer anywhere would call this “quick”.

36

Discovery to Containment
In the previous report, we presented this phase under the heading “Discovery to Mitigation.” Afterward, it was determined
that “Containment” provided better clarity as to our meaning. By containment, we essentially mean to “stop the bleeding”; it
does not refer to complete remediation of the root problem(s). Once a breach is contained, unauthorized access is cut-off
and information is no longer exposed.

As one can imagine, containing a data compromise situation as quickly as possible is paramount to the response process.
Unfortunately, this goal is rarely met. The majority of breaches required weeks or more to reign in and very few were contained
in less than a couple days. We believe insufficient preparedness to be the crux of the issue—many organizations are simply
not ready to respond when a crisis occurs. As a testimony to this, findings pertaining to response practices in place at the
time of the incident show that a large proportion of organizations were not adequately prepared to handle an incident.
Specific findings concerning the maturity of incident response practices are presented in the “Discovery and Response”
section that follows.

Discovery and Response
In addition to time-to-discovery and time-to-containment statistics, information surrounding the method of discovery,
detection capabilities, and incident response practices can help develop a clearer picture around what contributes to the
frequency and severity of breaches. As this topic is so critical and yet the findings from our previous report so deplorable, new
metrics were added for 2008 to allow further study. We begin with the method of discovery.

Discovery Methods
In last year’s study, we reviewed how organizations became aware that a breach had occurred. 2008 investigations reveal no
significant differences in the method of discovery; the vast majority of breaches continue to be discovered by a third party.
Some will notice a divergence in Figure 32 with respect to the “Notification by Third Party” category presented in the previous
report. Because this discovery method was and is so over-represented, we separated fraud detection from other types and
reasons of third-party notification. From this view, it is clear that breaches go unnoticed until a sufficient number of fraudulent
transactions enable a third party to isolate the point of compromise.

Employees happening upon breaches during their normal work activities remain a distant second (to the combined third-
party detection or notification categories) at 13 percent, followed closely by unusual system behavior at 11 percent. These
two methods, while unplanned, seem to be an organization’s best chance of discovering a breach on their own. That’s hard
to believe, but it’s also hard to argue with five years of data that all reveal a similar finding.

A simplified view of breach discovery methods is given in Figure 36, which concisely contrasts the regularity of third party and
internal discovery. As with prior years, “active” measures (referring to those that are specifically designed for detection) taken
by the organization detect only a small proportion of breaches. It is clear that breach discovery remains a largely passive
endeavor in 2008.

and look for the “Discovery and Response” post.

37

Figure 32. Breach discovery methods by percent of breaches

Third-party detection of fraudulent activity 55%

Third-party notification (not due to fraud) 15%

employee discovery during work activities 13%

unusual system behavior or performance 11%

event monitoring or log analysis 6%

Routine internal audit 2%

Blackmail or extortion 2%

Confession or brag by perpetrator 1%

Routine third-party audit 0%

The apparent ineffectiveness of event monitoring and log analysis continues to be somewhat of an enigma. The opportunity
for detection is there; investigators noted that 66 percent of victims had sufficient evidence available within their logs to
discover the breach had they been more diligent in analyzing such resources. Though lower than in previous years (it was 82
percent from 2004 to 2007), this finding still suggests that realized effectiveness remains much lower than potential
effectiveness. Of course, the extent of this underperformance depends
greatly upon the deployment of these technologies. We examine this in the
next section. Figure 33. Breach discovery
methods, simplified

Utilization of Detective Controls 7%
Internal
The information presented in the previous section raises obvious questions Active
concerning the prevalence and operational maturity of detection-oriented
controls. To help find answers, investigators gathered information on the 24%
utilization of these controls among organizations within our 2008 caseload. Internal
The findings are presented in Figure 34. passive

The data suggest that, beyond basic system and device logs, the utilization of 69%
Third
detective controls among breach victims is relatively low. If your thought Party
process is anything akin to ours, your eye went immediately to the 30 percent

38

attributed to intrusion detection systems and you muttered Figure 34. Detective controls by percent of
breach victims
“Gee—that seems low.” The temptation is to write this off as a
71%
byproduct of a sample that is obviously comprised of small
organizations and immature security programs. This is not the
case and disregarding these results as non-reflective of larger
enterprises would be a mistake. Though still unexpectedly low,
there are several valid reasons that help explain this result.

First, a number of large organizations that one would fully
expect to have intrusion detection systems (IDS) did not. Others
had deployed IDS but had not activated them. Still more (about 30%
12 percent) were monitoring network activity with IDS but not in
the locations or assets involved in the breach. In other cases, we 20%
were simply unable to confirm the existence of IDS and other
13%
detective capabilities (for various reasons). Finally, it is worth 11%
noting that some organizations represented in the study did not
have prescriptive compliance requirements for event detection.
System Intrusion Automated SeIM log
Of the five breaches that were detected by event monitoring Device Detection log (Info and Review
and log analysis, the IR team noted that three collected syslogs logs System Analysis event Mgt.) process

and ran IDSs, one collected syslogs and regularly reviewed them,
and one employed all detective controls listed in Figure 38.

Figure 35. Incident response practices by percent of breach victims

Incident Response Team 32%

Incident Response plan 28%

Incident Awareness and Response Training 25%

Regular Mock-Incident Testing 2%

39

Incident Response practices

During 2008 cases, investigators also noted the existence of the Incident Response practices shown in Figure 35. Overall, the
results suggest that, in addition to low detective capability, most organizations are ill-equipped to adequately respond to
breaches when they are discovered. We found it especially surprising that only 28 percent of victims had an incident response
plan in place.

Anti-Forensics
When criminals decide to break into corporate networks and steal data, getting caught is not typically part of their plan.
Many will attempt to muddy the crime scene and remove evidence of their actions to foil post-incident investigations. Such
activity is referred to as anti-forensics and is at least partially responsible for the breach discovery and response struggles
discussed in the previous section.

Investigators found signs of anti-forensics in over one-third of cases in 2008. The most common forms of anti-forensics
measures observed in the field are Data Wiping, Data Hiding, and Data Corruption. The prevalence of these among
investigations conducted in the last year is shown in Figure 39. The pervasiveness of Data Wiping, which includes removal
and deletion, comes as no surprise. Many commercial off-the-shelf products are available that perform this function (there
are many legitimate uses for wiping data) as well as freeware utilities. With respect to Data Hiding, the use of steganography
has remained relatively rare and flat year-over-year. On the other
hand, the use of encryption for the purposes of Data Hiding has
Figure 36. Types of anti-forensics encountered
by percent of breaches risen by almost 10 percent. Where Data Corruption was observed,
31% it was mostly manifested as log tampering.

Intuitively, the breaches in which anti-forensics were utilized
tended to be larger in size and longer in duration. For many of
the smaller “smash and grab” attacks, destroying evidence is
simply not worth the effort (plus those attacks are often
perpetrated by attackers of lower skill who are ignorant of anti-
forensic measures). In this sense, the use of anti-forensics is
progressing beyond its roots as a defensive tactic used to avoid
9%
prosecution and is increasingly employed in an offensive capacity
to expand and extend the attacker’s ability to compromise data.
3%

Data Data Data
Wiping hiding Corruption

and look for the “Anti-Forensics” post.

40

Payment Card Industry Data Security Standard
A great deal of public discussion has taken place regarding the effectiveness of regulations and control guidelines to prevent
breaches, but empirical study of the topic remains scarce. While it is not the purpose of this report to provide this much-
needed analysis, our 2008 data set does offer some information concerning incidents where the organization is required to
comply with the Payment Card Industry Data Security Standard (PCI DSS).

PCI DSS is a set of control requirements created to help
organizations protect cardholder information. Based on Figure 37. pCI compliance status based on last assessment
the representation of financial institutions and retailers by percent of breach victims

in our data set, it is no surprise that the large majority of
organizations within our caseload are subject to the
19%
requirements set forth within PCI DSS. Because these
Claimed/Found
were data compromise cases, an obvious question Compliant
arises as to the compliance status of the breach victims.
The answer to this question is given in Figure 40 but
additional clarification is needed to ensure proper
81%
interpretation.
NOT
Compliant
Over three-quarters of organizations suffering payment
card breaches within our caseload were found not
compliant with PCI DSS or had never been audited. This
status was not determined by our Investigative Response
team but rather by the victim’s attestation or Qualified
Security Assessor (QSA). Far more interesting than these
non-compliant organizations, however, are the 19 percent that had been found compliant during their last assessment. Does
the fact that these organizations were deemed compliant by their QSA and yet still suffered an incident mean that PCI DSS
was ineffective at preventing those breaches? Our findings do not support such a conclusion.

Due to the point-in-time nature of PCI assessments, it is possible that an organization deemed compliant at its last audit may
not still be compliant at the time of the breach. Furthermore, PCI compliance is not an absolute guarantee against breaches
nor is the assessment process always consistent. Whatever the reason, post-breach reviews conducted by our IR team reveal
some very interesting findings.

At the culmination of a forensic engagement, the investigator working the case performs a review of which PCI DSS
requirements were and were not in place (and adequate) at the time of the breach. The results of this assessment are recorded

and look for the “PCI DSS” post.

41

in the PCI Requirements Matrix, appended to the case report, and then conveyed to the relevant payment card brands. This
exercise is not an official PCI certification audit and it does nothing to uphold or overrule the victim’s compliance status. It
does, however, provide useful insight into which requirements tend to be deficient relative to organizations that suffer data
breaches. The aggregate results of post-breach PCI reviews conducted during 2008 are tallied in Table 10.

Table10. Results of post-breach pCI DSS reviews conducted by Verizon Business IR. Values represent the percentage of organizations
for which each requirement was found to be in place.

Build and Maintain a Secure Network Compliance
Requirement 1: Install and maintain a firewall configuration to protect data. 30%

Requirement 2: Do not use vendor-supplied defaults for system passwords and other security parameters. 49%

Protect Cardholder Data

Requirement 3: protect stored data. 11%

Requirement 4: encrypt transmission of cardholder data and sensitive information across public networks. 68%

Maintain a Vulnerability Management Program

Requirement 5: use and regularly update AV. 62%

Requirement 6: Develop and maintain secure systems and applications. 5%

Implement Strong Access Control Measures

Requirement 7: Restrict access to data by business need-to-know. 24%

Requirement 8: Assign a unique ID to each person with computer access. 19%

Requirement 9: Restrict physical access to cardholder data. 43%

Regularly Monitor and Test Networks

Requirement 10: Track and monitor all access to network resources and cardholder data. 5%

Requirement 11: Regularly test security systems and processes. 14%

Maintain an Information Security Policy

Requirement 12: Maintain a policy that addresses information security. 14%

42

When reviewing the percentages for each requirement above, several very interesting statistics pop out. Requirements 3, 6,
and 10—which many organizations complain are the most onerous—are indeed the least compliant across our caseload.
When one considers the prevalence of unnecessary and/or unknown data stores, frequency of SQL injection attacks, and
lengthy compromise-to-discovery periods discussed extensively in this (and our last) report, this finding is hardly surprising.
This trio of deficiencies factored heavily into many of the largest breaches investigated by our team over the past five years.
Unfortunately, these statistics—and those discussed earlier regarding the low utilization of discovery and detective controls—
suggest that attacks exploiting these areas will continue to be a challenge for the foreseeable future.

In other words, the typical organization had met
less than a third of the requirements in PCI DSS.
Some fared much better, some much worse, but
the point made by the data before us is this: these
breaches, in general, did not occur in organizations
that were highly compliant with PCI DSS.
On the other end of the spectrum, encrypting data over public networks and antivirus software exhibit fairly high levels of
compliance (68 percent and 62 percent respectively). Unfortunately, they have little to do with common attack modalities.
As discussed earlier in this report, information is either captured off private networks or within systems and AV software is of
little effect against customized malware installed on a system under the control of a remote attacker. Nevertheless, antivirus
software certainly plays a major role in enterprise defenses and the fact that 40 percent of victims were not using it (or not
updating it) is quite surprising.

Incredibly, 51 percent of the victims were using vendor default passwords on systems that handle sensitive data. In over 80%
of our cases, the victim organizations used shared accounts for system access and did not assign unique IDs to users. That the
low adherence to these requirements played a role in the numerous breaches attributed to default and shared credentials
used for third-party remote administration services is indisputable.

While the requirement-centric view of this information is interesting, the organization-centric view makes a compelling point
as well. Our unofficial post-breach PCI reviews found the average compliance rate across victims in our caseload to be 29
percent of the 12 requirements. In other words, the typical organization had met less than a third of the requirements in
PCI DSS. Some fared much better, some much worse but the point made by the data before us is this: these breaches, in
general, did not occur in organizations that were highly compliant with PCI DSS.

43

Conclusions and Recommendations
The conclusion to our previous report began with the statistic that, in 87 percent of cases, investigators concluded that the
breach could have been avoided if reasonable security controls had been in place at the time of the incident. Some felt our use
of “reasonable” was not entirely clear (you can read our explanation here*), so we opted for a different route this time. That route,
however, arrived at the very same conclusion—just worded differently.

In 2008, investigators concluded that 87 percent of breaches could
Figure 38. Description of the effort and expense have been avoided through the implementation of simple or
of recommended preventative measures by
percent of breaches intermediate controls. All of these were the standard, run-of-the-mill
practices that we in the industry see and use everyday. Costly
controls (in terms of effort and expense) were recommended as the
most efficient and effective means of avoiding the breach in only 13
percent of cases. Furthermore, most of these—though costly—were
34%
Intermediate 53% standard security controls. Figure 39 provides a different classification
Simple and of these same controls.
Cheap
At the conclusion of the original DBIR, we made a number of
recommendations based directly on findings from our 2004 through
2007 breach investigations. Additionally, we called for a shift in the
13% data protection and incident response mentality that pervades the
Difficult and industry. This year, we would like to reiterate that call and recapitulate
expensive
those recommendations as well as provide additional guidance
based on our 2008 caseload.

An abbreviated version of recommendations from the 2008 DBIR follows. They are as relevant today as they were last year.

Align process with policy: Many organizations set security policies and procedures yet fail to implement them consistently.
Controls focused on accountability and ensuring that policies are carried out can be extremely effective in mitigating the risk of a
data breach.

Achieve essential, and then worry about excellent: We find that many organizations achieve very high levels of security
in numerous areas but neglect others. Criminals will almost always prefer the easier route. Identifying a set of essential
controls and ensuring their implementation across the organization without exception, and then moving on to more
advanced controls where needed is a superior strategy against real-world attacks.

Secure business partner connections: Basic partner-facing security measures as well as security assessments, contractual
agreements, and improved management of shared assets are all viewed as beneficial in managing partner-related risk.

*http://securityblog.verizonbusiness.com/2008/06/19/reasonable-controls/

44

Figure 39. Simple categorization of recommended preventive measures by percent of breaches

Configuration Change to exisiting Assets 40%

Alter exisiting practice 27%

Implement new practice 25%

Training and Awareness 16%

Alter or Redeploy exisiting Assets 11%

Deploy new Assets 11%

policy update 9%

Create a data retention plan: Clearly, knowing what information is present within the organization, its purpose within
the business model, where it flows, and where it resides is foundational to its protection. Where not necessitated by valid
business needs, a strong effort should be made to minimize the retention and replication of data.

Control data with transaction zones: Based on data discovery and classification processes, organizations should separate
different areas of risk into transaction zones. These zones allow for more comprehensive control implementations to
include but not be limited to stronger access control, logging, monitoring, and alerting.

Monitor event logs: All too often, evidence of events leading to breaches was available to the victim but this information
was neither noticed nor acted upon. Processes that provide sensible, efficient, and effective monitoring and response are
critical to protecting data.

Create an Incident Response Plan: If and when a breach is suspected to have occurred, the victim organization must be
ready to respond. An effective Incident Response Plan helps minimize the scale of a breach and ensures that evidence is
collected in the proper manner.

Increase awareness: Delivered effectively, training that educates employees about the risks of data compromise, their
role in prevention, and how to respond in the event of an incident can be an important line of defense and discovery.

45

Engage in mock incident testing: In order to operate efficiently, organizations should undergo routine IR training that
covers response strategies, threat identification, threat classification, process definition, proper evidence handling, and mock
scenarios.

In addition to these “oldies but goodies,” we offer the following new or expanded recommendations based on our analysis of
2008 data. As before, these recommendations are in no way a comprehensive strategy for enterprise data protection and
certainly cannot guarantee against compromise. Rather, they are a consequence of the experience gained through the
hundreds of data breach investigations and are intended to highlight specific problem areas common to many organizations
within our caseload.

Changing default credentials is key: More criminals breached corporate assets through default credentials than any
other single method in 2008. This certainly falls under the prior “achieve essential, and then worry about excellent”
recommendation, but the prevalence of this specific failure justifies its special mention. It is not as though the victims had
no idea that changing usernames and passwords was important; they just neglected to do so on a few assets (and the
intruders, of course, found those) or they assumed the third party responsible for managing them would do so. The
lesson? Don’t assume that your staff or your partner’s consistently follows through on all policies and procedures.

Avoid shared credentials: Another obvious yet frequently omitted and oft-exploited problem. Along with changing
default credentials, organizations should ensure that passwords are unique and not shared among users or used on
different systems. The use of shared credentials allowed quite a few breaches in 2008 to grow from a single point of
compromise to multiple incidents. This was especially problematic for assets managed by a third party.

User account review: This is related to the two prior recommendations but worth calling out individually. Default and
shared credentials, a growing rate (and large number of records lost) of breaches involving unknown privileges, and years
of experience lead us to believe in the value of reviewing user accounts on a regular basis. The review should consist of a
formal process to confirm that active accounts are valid, necessary, properly configured, and given appropriate (preferably
least) privileges.

Application testing and code review: SQL injection attacks, cross-site scripting, authentication bypass, and exploitation
of session variables contributed to nearly half of breaches attributed to hacking or network intrusion. It is no secret that
attackers are moving up the stack and targeting the application layer. Why don’t our defenses follow suit? As with
everything else, put out the fires first: even lightweight web application scanning and testing would have found most of
the problems that led to major breaches in the past year. Next, include regular reviews of architecture, privileges, and
source code. Incorporating a Security Development Life-Cycle (SDLC) approach for application development is
recommended as well. Finally, help your developers learn to appreciate and write more secure code.

Smarter patch management strategies: For every vulnerability exploited by hacking and malware attacks in 2008, the
patch necessary to prevent the breach had been available for at least six months prior to the incident. In fact, all but one
had been around for a year or more. While it may seem logical to conclude that organizations aren’t patching fast enough,
this is not the correct interpretation. All of these organizations had patch cycles well below the six month mark. The

46

problem throughout all five years of this study has far more to do with scope than speed. Organizations would find much
more value if they divert resources from patching ever-faster to patching more consistently and comprehensively.

Human resources termination procedures: Several breaches in the last year were the result of malicious activity on the
part of a recently terminated (or notified) employee. Each organization should have a comprehensive employee
termination policy that identifies parties responsible for the termination process, outlines procedures and checklists for
termination, ensures the retrieval of all organizational property, and—most importantly—establishes a process for quickly
disabling user accounts and removal of all access permissions.

Enable application logs and monitor them: Yes, “monitor event logs” is a recommendation we gave last year and have
already recapped above. However, we find that many organizations focus these logging efforts on network, operating
system, IDS, and firewall logs and neglect remote access services, web applications, databases, and other critical
applications. These can be a rich data set for detecting, minimizing, and investigating breaches.

Define “suspicious” and “anomalous” (then look for whatever “it” is): This is admittedly vague, but—in truth—
generalizing what this entails in order to prescribe something for everyone would counteract the point. During this
report, we have discussed extensively the increasingly targeted and sophisticated nature of a small but damaging subset
of attacks. These typically occur within organizations that process or store large quantities of data valued by the criminal
community; they are the quintessential Targets of Choice. Organizations falling under this classification should prepare to
defend against and—especially—detect very determined, well-funded, skilled, and targeted attacks. Consider the “big
breach” scenario highlighted in this report: The attacker gains access to the corporate network (potential warning sign
number one), explores around a while (sign #2) to find and breach (sign #3) sensitive systems, installs malware (sign #4) that
alters the state and processing of the system (sign #5). The malware stores data locally—often on a system that should not
be storing any data (sign #6)—and the criminal returns periodically (sign #7) to collect the data and often does so via
unusual ports (sign #8) and from known malicious IP addresses (sign #9).. Sound suspicious and anomalous to you? We
think so too. Discover what is critical, identify what constitutes normal behavior, and then set focused mechanisms in place
to look for and alert upon deviations from normality.

285 million records breached in a single year is a rather loud wake up call to an industry dedicated to protecting information. We
can’t afford to hit snooze and sleep in. Our task is not getting any easier; the sum total of information in the world grows
continually and permeates everything we do and everywhere we go. While the majority of attacks remain rather mundane (let’s
take care of those first, by the way), the bad guys are adapting to our current protection strategies and inventing new ways to
attain the data they value.

The good news is that we have more and more information at our disposal. A growing body of research aims to leverage real
data and real results to help us better perceive reality and ensure that we are focusing on the right things. Sure, there’s still a lot of
noise out there but signal appears to be getting stronger. We hope our contribution is considered to be more signal than noise.

47

About the Verizon Business Investigative Response Team
Security breaches and the compromise of sensitive information are a very real concern for organizations worldwide. When such
incidents are discovered, response is critical. The damage must be contained quickly, customer data protected, the root cause
found, and an accurate record of events produced for authorities. Furthermore, the investigation process must collect this
evidence without adversely affecting the integrity of the information assets involved in the crime.

The IR team has a wealth of experience and expertise, handling roughly 600 security breach and data compromise cases in the
last five years. This includes a substantial proportion of all publicly disclosed breaches as well as many that have never been
released. This caseload represents a large proportion of total known compromised records during this timeframe and many of
the largest breaches ever reported.

During such investigations, the team regularly interacts with governmental agencies and law enforcement personnel from
around the world to transition case evidence and set the stage for prosecution. In addition to security breach and data
compromise cases, the IR team provides services such as litigation support, eDiscovery, expert witness testimony, chain-of-
custody, mock-incident training, and incident response program development.

The expansive statistical data set generated through these activities offers an interesting glimpse into the trends surrounding
computer crime and data compromise.

yr puvsser vaqrpuvssenoyr

48

www.verizonbusiness.com
© 2009 Verizon. All Rights Reserved. MC13626 0409
Subject to Terms of use available at http://securityblog.verizonbusiness.com/disclaimer-2/.
The Verizon and Verizon Business names and logos and all other names, logos, and slogans identifying Verizon’s products
and services are trademarks and service marks or registered trademarks and service marks of Verizon Trademark Services
llC or its affiliates in the united States and/or other countries. All other trademarks and service marks are the property
50 of their respective owners.

2009 Databreach Report

More Related Content

What's hot (12)

Viewers also liked (16)

Similar to 2009 Databreach Report (20)

More from GuardEra Access Solutions, Inc. (13)

Recently uploaded (20)

2009 Databreach Report