
Towards Next-Generation Botnets
Ralf Hund Matthias Hamann Thorsten Holz
Laboratory for Dependable Distributed Systems
University of Mannheim, Germany
Abstract
In this paper, we introduce the design of an ad-
vanced bot called Rambot that is based on the weak-
nesses we found when tracking a diverse set of botnets
over a period of several months. The main features of
this bot are peer-to-peer communication, strong cryp-
tography, a credit-point system to build bilateral trust
amongst bots, and a proof-of-work scheme to protect
against potential attacks. The goal of this work is to
increase the understanding of more advanced botnet
designs, such that more efficient detection and mitiga-
tion systems can be developed in the future.
1 Introduction
A bot is a computer program installed on a com-
promised machine which allows an attacker to execute
arbitrary commands on the infected machine. Botnets,
i.e., networks of such bots under a common control in-
frastructure, are one of the main problems in today’s
Internet since botnets are often used to carry out Dis-
tributed Denial-of-Service (DDoS) attacks, send a large
amount of spam or phishing mails, and other nefarious
purposes [4, 15, 22].
In response to this problem, researchers developed
a diverse set of techniques and tools to either detect
or mitigate botnets [2, 8, 11, 12, 18]. Unfortunately,
most of these techniques focus on botnets with a central
server used for Command & Control (C&C): in these
scenarios, the bots connect to a central server from
which they receive the commands they are supposed
to execute. Mitigation of such botnets can be achieved
by locating the central server and then shutting it
down [8]. However, in the recent months, the first bot-
nets that use a peer-to-peer (P2P) based communica-
tion channel appeared [6, 10, 21, 26, 27]. Dealing with
such botnets is a challenging task, still some techniques
to track these botnets have been proposed [14].
In the near future, we expect the emergence of more
advanced botnets that can defeat the current defense
mechanisms. In this paper, we present the design
of one such advanced botnet named Rambot, a next-
generation botnet whose design is based on the lessons
we learned when tracking a diverse set of botnets over
a period of several months. We address all the weak-
nesses of current botnets we noticed and construct a
botnet that is hard to track and shut down. While
some of our design decisions seem to be obvious, e.g.,
using strong cryptography, (almost) no current botnet
uses them. Furthermore, some aspects of Rambot’s
design are countermeasures against new techniques to
track botnets, e.g., proof-of-work schemes to counter
crawling of the whole botnet.
One basic design aspect of the botnet developed in
this paper is P2P-based communication. While de-
centralized botnets overcome many of the problems
of botnets with a central server, e.g., there is no sin-
gle point of failure, also several distinct weak points
against someone who wants to shut down the network
arise. In a central botnet, many security related re-
quirements, such as, for example, authentication of
botmaster commands, are guaranteed as long as the
central server has not been compromised. While this
has been proved to be a reasonable assumption, things
are more complicated concerning a decentralized net-
work since each bot carries additional responsibilities.
We illustrate these implications throughout the paper
and subsequently offer solutions to such problems. In
summary, Rambot is an example of the threats and
challenges we expect in the near future.
The area of botnets faces – similar to other fields –
an arms race between botmasters and defenders. To
keep up with latest developments, researchers need to
continue to improve detection and mitigation methods
and investigate new techniques used by botmasters.
The goal of our work is to increase the understand-
ing of more advanced botnet designs. We anticipate
that this paper ultimately leads to the development of
new, sophisticated techniques, which will help to fend
off arising threats.
1