A Survey of Botnet Technology and Defenses
Michael Bailey, Evan Cooke, Farnam Jahanian, Yunjing Xu
University of Michigan
Ann Arbor, Michigan
{mibailey, emcooke, farnam, yunjing}@umich.edu
Manish Karir
Merit Network, Inc.
Ann Arbor, Michigan
mkarir@merit.edu
Abstract
Global Internet threats have undergone a profound
transformation from attacks designed solely to disable in-
frastructure to those that also target people and organiza-
tions. At the center of many of these attacks are collections
of compromised computers, or Botnets, remotely controlled
by the attackers, and whose members are located in homes,
schools, businesses, and governments around the world [6].
In this survey paper we provide a brief look at how existing
botnet research, the evolution and future of botnets, as well
as the goals and visibility of today’s networks intersect to
inform the field of botnet technology and defense.
1 Introduction
Global Internet threats are undergoing a profound trans-
formation from attacks designed solely to disable infras-
tructure to those that also target people and organizations.
This alarming new class of attacks directly impacts the day-
to-day lives of millions of people and endangers businesses
and governments around the world. For example, computer
users are assailed with spyware that snoops on confidential
information, spam that floods email accounts, and phishing
scams that steal identities.
At the center of many of these attacks is a large pool of
compromised computers located in homes, schools, busi-
nesses, and governments around the world. Attackers use
these zombies as anonymous proxies to hide their real iden-
tities and amplify their attacks. Bot software enables an
operator to remotely control each system and group them
together to form what is commonly referred to as a zom-
bie army or botnet [6]. The scope of the botnet problem is
difficult to quantify, as the highly covert nature of bots and
botnets makes them difficult to identify and even harder to
measure. Nevertheless, CERT has identified botnets with
more than 100,000 members, and almost 1 million bot in-
fected hosts have been reported [19].
In this paper, we provide a survey of current botnet tech-
nology and defense by exploring the intersection between
existing botnet research, the evolution of botnets them-
selves, and the goals and perspectives of various types of
networks. In section 2, we provide a brief overview of bot-
nets to highlight the invariant nature of their behavior in
various phases of their life-cycle. Then, in section 3, we
describe how different kinds of networks have access to dif-
ferent types of visibility and this has a strong impact on the
effectiveness of any botnet detection mechanism. Next, in
section 4, we provide a comprehensive overview of the var-
ious botnet detection techniques that have been been pro-
posed. Finally, in section 5, we summarize our survey and
suggest future directions.
2 Understanding Botnets
In many respects, the bots found in the wild today are a
hybrid of previous threats. They can propagate like worms,
hide from detection like many viruses, attack like many
stand-alone tools, and have an integrated command and
control system. Even more concerning, the construction of
bots is now very much a cooperative effort. An example is
the source code of SDBot, which contains comments from
many different authors. The result is a proliferation of dif-
ferent bot variants. A recent Microsoft survey found more
than 43,000 new variants of backdoor trojans and bots dur-
ing the first half of 2006 [20].
2.1 Propagation and Compromise
One core problem for botnet attackers is how to get bots
onto victim computers. Because very few users would ac-
tually agree to have their computers used to conduct packet
floods, attackers surreptitiously install their malicious soft-
ware. This process of getting malicious software on vic-
tim’s hosts has evolved significantly over time. One change
that happened a few years ago is the shift from a single prop-
agation vector, that might have required a manual installa-
tion process by the attacker, to multiple automated propaga-
tion vectors. For example, The Slammer worm used a single