Proceedings of CCIS2014
DETECTING THE RUN TIME ATTACKS IN THE CLOUD
WITH AN EVIDENCE COLLECTION BASED APPROACH
Jie lin
1,3
, Chuanyi Liu
2,3*
, Zhichun Ning
3,4
, Binxing Fang
1,3
1
School of Computer Science, Beijing University of Posts and Telecommunications, Beijing, China
2
School of Software Engineering, Beijing University of Posts and Telecommunications, Beijing, China
3
Key Laboratory of Trustworthy Distributed Computing and Service (BUPT), Ministry of Education, Beijing, China
4
School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing, China
* Corresponding Author: cy-liu04@mails.tsinghua.edu.cn
Abstract: As the impacts of security threats are
amplified and spread quickly in the cloud, especially
with the attribute of large-scale virtualization. Most of
researches focus on monitoring the network traffic.
However, these methods don’t check how the attack
happened and what damage the attack caused for the
virtual infrastructure. This paper presents a logging
method to analyze the potential attacks out of VM and a
prototype was implemented. The experimental results
show that this method can detect the real world attacks
effectively and the performance overhead is acceptable.
Keywords: Virtual machine; Cloud; Security; Control
1 Introduction
With the typical characteristics of cloud computing, such
as large-scale resource pooling, many virtual machines
generated by the same template, multi-tenants, so once
the attackers control the cloud, they may utilize the
cloud to launch large-scale attacks, e.g. sending spam
emails, performing Distributed Denial of Service (DDoS)
attack [1]. How to effectively control and manage the
cloud and prevent cloud becoming a facility for the
attackers is an urgent problem.
To detect the behavior that utilizing the cloud to perform
attacks, most researchers proposed some methods to
analyze the network traffic [1, 2]. But these methods
doesn’t analyze how the cloud is controlled by the
attackers. There needs a method to efficiently analyzing
the process of attack, that is logging and auditing the
events happened in the virtual machine (VM). In the
virtualization, the main challenges to effectively collect
and analyze the evidence outside the VM are as follows.
z The isolation between target system and collection
software. Installing collection software in the VM
exposures the software to the attacker. Putting the
collection software out of VM can reduce the risk
of being attacked.
z Semantic gap [3] between high level languages and
computer architecture, including VM. The data got
from VM is low level binary. Reconstruction of the
semantic needs to be solved.
z The requirement for fine-grained gathering method
and low performance overhead. There are many
methods intercepting the instructions. But they
don’t fit for the real time collection because of the
low efficient performance.
This paper presents a method to collect the evidence
happened in the VM. This method isolates the target
system and collection module. The evidence will be
analyzed for potential attacks. The main contributions of
this paper include:
z A method outside the VM to log various
information was proposed, including event,
memory, vCPU state. This paper analyzes the
runtime attacks essence, and gives the
corresponding countermeasure.
z A prototype for detecting the runtime attacks is
implemented. This paper demonstrates the detailed
architecture design. The architecture includes the
logging part and the analysis part.
z Using the real world attack to evaluate the
effectiveness of proposed method, and the
performance was conducted.
The rest of the paper is organized as follows. Section 2
covers related work. Section 3 presents the architecture
design. In section 4, the evaluations are provided,
including effectiveness and performance. The
conclusions are presented in section 5.
2 Related work
Researchers have proposed many approaches to detect
the potential threat. In this section, the discussions are
mainly focused on the work related to virtualization.
Integrity measurement. This method is used to check
whether the program is tampered with. The concept of
system integrity was proposed to protect the system
against unauthorized user access [4]. Some approaches
try to protect the system according to the integrity
attribute. IMA [5] makes measurement of executable
code and data file (e.g., modules or configuration files),
and stores the hash values of them into a measurement
list. Modification to the executable code and data file
will be recorded. To separate measurement agent and
measurement target, HIMA [6] locates in the hypervisor
which provides the strong isolation. HIMA uses active
monitoring to solve the Time-of-Check-to-Time-of-Use
(TOCTTOU) [4] consistency. But above integrity
measurement methods may fail in some runtime attacks,
which will be explained in this paper.
Virtual machine introspection. Virtual machine
introspection [7] is an approach to inspect VM from the
____________________________________
978-1-4799-4719-5 /14/$31.00 ©2014 IEEE