A Dynamic Ddos Protection Mechanism for WLAN based on SDS
Architecture
Zhenyu Wang
1, 2
, Heng He
1, 2
, Yan Hu
1, 2
, Ji Zhang
1, 2
, Wei Xia
1, 2
1
School of Computer Science and Technology, Wuhan University of Science and Technology, Wuhan, Hubei 430065, China
2
Hubei Province Key Laboratory of Intelligent Information Processing and Real-time Industrial System, Wuhan University
of Science and Technology, Wuhan, Hubei 430065, China
374487006@qq.com, heheng@wust.edu.cn, 1540823670@qq.com, 1143828107@qq.com, xiawei137hao@sina.com
Keywords: Distributed Denial of Service, Software Defined Networking, Protection Mechanism, Wireless Local Area
Network.
Abstract: The impact of distributed denial of service (DDoS) attacks has become more and more serious and
widespread in wireless local area network (WLAN). Traditional DDoS protection mechanisms become less
reliable and cannot easily adapt to the diverse types of DDoS attacks. Meanwhile, the emergence of
software defined networking (SDN) has provided a new solution to solve the security problem in WLAN. In
this paper, we propose a dynamic DDoS protection mechanism for WLAN based on software defined
security, which is a branch of SDN architecture in the network security. When outer-net data flow streams
into the network, the mechanism can judge the credibility of the flow by its self-detection function, and then
it will deploy different security strategies to handle the data flow according to the credibility before server
responds to it. The analysis and experiment show that compared with traditional DDoS protection
mechanisms, the proposed mechanism is a priori detection method, and is more flexible and efficient.
1 INTRODUCTION
With the rapid technological development of the
Wireless Local Area Network (WLAN), the user
requirements for security and trust of WLAN are
also increasing significantly. Because WLAN has its
own unique security threats, for instance, the IEEE
802.11 series standards and WEP/WPA encryption
protocol have obvious defects, the WLAN security
issues have become increasingly prominent.
Distributed Denial of Service (DDoS) attack is a
kind of typical attack in WLAN and the DDoS
protection mechanism of WLAN has become a hot
topic in the research field of network security (Yu et
al., 2011), (Tupakula et al., 2011).
DDoS attack is an attempt to make a machine or
network resource unavailable to its intended users.
Although the means to carry out, the motives for,
and the targets of a DDoS attack vary, it generally
consists of efforts to temporarily or indefinitely
interrupt or suspend services of a host connected to
the Internet. DDoS attack in WLAN is mainly
divided into the following types: bandwidth
consumption attacks, resource consumption attacks,
identity fraud attacks, interference type attacks, etc.
Bandwidth consumption attacks include ICMP
flooding, UDP flooding, etc. Resource consumption
attacks include SYN flooding, auth flooding, etc.
Identity fraud attacks include deauth flooding,
association flooding, disassociation flooding, etc.
Interference type attacks include teardrop, RF
Jamming, etc. Based on these attacking types, there
are lots of protection solutions proposed in recent
years, such as route-based filtering (Park, 2003),
packets analysis (Zhang et al., 2009) anomaly
detection (Thatte et al., 2011), etc. However, most of
the solutions are efficient for certain type of attack,
but work little to other types of attacks, while DDoS
attacks are presenting more diverse and complicated
trend in WLAN. Therefore, it’s necessary to propose
a dynamic solution by integrating the previous ones,
which is flexible to deal with most of the types of
attacks.
Nowadays, network technology becomes more
mature and gradually develops into a new network
architecture, Software Defined Networking (SDN)
(Thomas and Ken, 2014). It is an emerging
41
He H., Xia W., Hu Y., Wang Z. and Zhang J.
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture.
DOI: 10.5220/0006443400410047
In ISME 2016 - Information Science and Management Engineering IV (ISME 2016), pages 41-47
ISBN: 978-989-758-208-0
Copyright
c
2016 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
41
architecture purporting to be dynamic, manageable,
cost-effective, adaptable, and seeking to be suitable
for the high-bandwidth, dynamic nature of today's
applications. The SDN architecture decouples
network control and forwarding functions, enabling
network control to be directly programmable and the
underlying infrastructure to be abstracted from
applications and network services. OpenFlow (OF)
(Lei, 2013) protocol is a foundational element for
building SDN solutions. OF separates the control
plane and data plane of network equipment, so as to
realize the flexible control of network traffic and
provide a good platform for the core network and
innovative applications. As a result, SDN offers
more new possibilities to solve network security
problems, including DDoS attacks. Software
Defined Security (SDS) is a branch of SDN
architecture in the network security, and it achieves
the separation and reconstruction of the data surface
and control surface, realizing modularity,
servitization and reusability.
In this paper, based on SDS architecture and
existing approaches, we propose a Dynamic DDoS
Protection Mechanism for WLAN, namely DDPM,
to solve the problem of diverse and complicated
DDoS attacks in WLAN effectively. According to
the types of DDoS attacks, DDPM deploys different
security strategies for the underlying network.
When outer-net data flow streams into the SDN
network, the system can judge the credibility of the
flow by its self-detection function, and then it will
make decision to handle the data flow according to
the credibility before server responds to it.
Compared with traditional DDoS protection
mechanisms, DDPM is a priori detection method,
which is more flexible and efficient.
2 DYNAMIC PROTECTION
MECHANISM BASED ON SDS
2.1 Architecture
DDPM inherits three main features of SDN
framework: centralized control, open interface and
virtualized network (Lei, 2013). The separation of
the data plane and the control plane atomizes the
functions and divides the system into five service
modules, which provide northern interface for the
invocation by higher layer. Meanwhile the
virtualized network shields the realization of devices
and thus reduces the difficulty of deployment.
Figure 1 shows the architecture of DDPM.
In Figure 1, DDPM is divided into five function
modules: Threat Detection module (TD), Credit
Evaluation module (CE), State Table module (ST),
Core Strategy module (CS) and Traffic
Identification module (TI).
Infrastructure
Layer
Thre at
Detection
module
Credit
Evaluation
module
Stat e
Table
module
Core
Strategy
module
Controller:
Floodlight,
Beacon,
Nox,
Traffic
Identificati
on module
Service
Layer
Control
Layer
SDN controller
OFSwitc h OFSwitch
OFRouter
Figure 1: Architecture of DDPM.
On the Infrastructure Layer, OFSwitch and
OFRouter, which are deployed in the SDN network,
maintain flow tables, device status and other
important information. When data flow streams into
the SDN network, these devices will specify
its action to forward or to discard.
On the Control Layer, SDN controller maintains
the underlying network topology, manages network
information, issues forwarding strategy and provides
northern interface to the higher layer. More
specifically, TI, which is deployed in the SDN
controller, processes the data flow that Infrastructure
Layer could not identify and then delivers the
underlying network information to Service Layer.
After receiving the developed strategy from Service
Layer, SDN controller transfers the strategy into
flow tables that specify data flow’s action and status
tables that maintain devices’ status.
Service Layer contains concrete implement of
DDPM. Firstly, TD detects the current status of
network according to the underlying network
information. Secondly, CE evaluates the credit level
of the data flow and preserves the values in the ST.
Actually, these three modules associated with each
other. Finally, CS will develop the newest strategy
according to the information from previous modules
if the system has detected the change of current
network status, and issue this strategy to the Control
Layer. Figure 2 shows the execution flow of DDPM.
2.2 Implement of the Modules
TI, as the function module of Control Layer,
provides intermediate hub for the Service Layer and
ISME 2016 - Information Science and Management Engineering IV
42
ISME 2016 - International Conference on Information System and Management Engineering
42
Infrastructure Layer. Data flow will be matched to
the flow tables, when it streams into the access layer
switch. After failing to match any flow tables, the
access layer switch will package this flow into a
Packet_In message and send it to the SDN
controller. Then, the controller will send the
message to TI for processing. First, TI will cluster
data flows’ information (Ramos et al., 2008), (Lee et
al., 2008) such as MAC, IP, data packet type and
port and so on. Figure 3 (1) shows the discrete
model of flows, with each record as a tuple. Figure 3
(2) shows the aggregation model of the flows. It
clusters source IP and integrates the same network
segment into one record. Compared to the discrete
model, aggregation model can greatly reduce the
number of records and the scale of flow table, which
greatly reduces the load between controller and
switch.
start
Outer data flow
streams into network node
Whether it match
flow table?
Identify and
classify flow
Add status table
information
Initial credit value
Issue flow table by
controller
Update status table
information
Check according to
credit level and
status table
Update credit value
Detect threat
level of flow
Yes
No
Select strategy 1 Select strategy N
1 N
Is it different
from original
strategy?
Issue strategy by
controller
Modify flow table
and network settings
end
No
Yes
Figure 2: Execution flow of DDPM.
After TI identifies the unrecorded flows, ST will
store each result in the library as one record. Once
the records are set up, ST will update the records
according to the information of the flow tables and
OFSwitch’s status which is obtained by SDN
controller as the same type of flows stream into the
switch. Every time ST updates its records, TD will
do once self-similarity detection (Xiang, 2004),
which aims to detect the change rate of information
including bandwidth, sending rate of packet,
amounts of flow tables and other important indexes.
First, detection of bandwidth change rate can
monitor the SDN network loads, so it can detect the
bandwidth consumption attack effectively. Second,
aggregating the flow tables based on source network
segment and calculating its matching frequency can
reflect DDoS attacks’ strength effectively, because
most cases of the sources of attacks come from a
handful of network even the forged sources.
However, given the trend of diverse attacks, TD also
detects the change rate of the amounts of flow
tables. When attacks come from many sources, the
scale of flow tables will increase dramatically.
CE, based on the results of TD’s detection,
evaluates each data flow’s credit. It divides the
credit into five levels: safety, alert, emergency,
danger and destruction, by which CS develops five
different strategies. Every time the certain flow is
over each level’s threshold, SDN controller will
send the packaged raw data flow to TD which will
analyze the data flow and separate the flow into two
parts. One part is the same with previous flow, while
another is the definite one over the threshold.
More specifically, after TI identifies the new data
flow, CE will mark the data flow as the default value:
safety. And then CS develops and issues a
forwarding strategy that the data flow selects the
optimal path to forward according to Dijkstra
algorithm. With the increasing sending rate of
certain flow over the alert-credit threshold, CE
marks it for “alert” and then CS changes this flow’s
forwarding strategy to select another lighter-load
forwarding path. By this way, it can reduce the load
of certain link and best utilize available bandwidth.
Furthermore, if the certain data flow is over the
emergency-credit threshold, CE will mark it for
“emergency” and then changes the strategy to drop
the certain packet type of the flow. This flow is
actually based on series of flows which comes from
different source MAC/ IP but same network segment.
As a result, which flow the system is going to drop
is the flow with definitized source MAC/IP and
packet type, while other flows within the same flow
cluster will not be limited. Once certain flow is over
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture
43
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture
43
Source Network Packet Destination Destination
IP Segment Type IP Port
Source Packet Destination Destination
IP Type IP Port
D1
D1
D1
D1
D1
T1
T2
T5
T4
T3
S1
S2
S3
S5
S4
P1
P2
P3
P4
P5
T3
T1
T2
N1
N3
N2
S1
S2
S3
S5
S4
D1
D2
D3
P1
P2
P3
1
2
Figure 3: Discrete and aggregation model of data flows.
danger-credit threshold, CS will change the strategy
to drop all this source-based flows. The
communication between server and normal users
will not be affected until any flows exceed the
destruction-credit threshold. At the time, CE marks
the flow for “destruction” and informs CS to change
the strategy to limit the receive rate of certain port
that this data flow streams into through issuing the
status-change command to the OFSwitch. There is
no doubt that all flows from this port will be affected
by limiting the import of the OFSwitch.
The process of DDPM’s protection is dynamic. It
offers several different strategies to handle different
intensities of data forwarding. Once certain flow
recovers to lower credit level, CS also needs to
change the strategy to temperate one. In addition,
with higher level of the flows’ credit, there are larger
scale of flow tables and higher frequency of data
receiving. As a result, it’s an elastic, real-time and
reliable mechanism but also has much higher
requirements of the performance of hardware.
3 ANALYSIS AND EVALUATION
It’s SDN network that makes this elastic and
dynamic mechanism possible. The characteristic of
service atomization makes DDPM easy to extend
new function modules to improve its reliability.
Compared to single and static traditional solutions,
DDPM is more like a set containing different
solutions. Based on different detection indexes, it
provides several detection results. Therefore, it’s
more reliable for DDPM because while one index
might be not relative to certain type of attack,
another index could be correlative to it.
On the way of protection, DDPM provides five-
level strategies, which means that all situations have
been divided into five different intensities of attacks.
With the increasing intensity of data transmission
rate, the system develops the stricter strategy. More
importantly, the data flow sending to server will be
detected as long as it streams into the SDN network.
In order words, DDPM can resist the potential DDoS
attack before server is under attack. Apparently, it’s
more efficient and safer to protect server from DDoS
attack than the traditional solutions which are based
on analysis of packets that server has received.
To build a SDN network requires only
OFSwitch, OFRouter and SDN controller because
DDPM is accomplished primarily in software and
does not need any other expensive protective
equipments. On the other hand, it also requires the
higher performance of these network equipments.
Under the enormous amounts of data flows,
forwarding devices need to match quantities of flow
tables simultaneously and SDN controller needs to
process large numbers of unrecognized data flows.
Moreover, it is difficult to ensure the reliability of
single controller within a large scale network.
However, the SDN technique is developing rapidly
and the collaboration among multiple controllers
will be well achieved in the near future. Above all,
the comprehensive comparison is shown in Table 1.
Table 1: Comprehensive comparison.
Solutions Traditional mechanism DDPM
Scalability Low High
Reliability Low High
Flexibility Low High
Protective method Single Multiple
ISME 2016 - Information Science and Management Engineering IV
44
ISME 2016 - International Conference on Information System and Management Engineering
44
30M
30M
30M
30M
Network 1
Network 2
Network 3
Server
SDN controller
OFSwitch 1 OFSwitch 2
OFSwitch 3
OFSwitch 4
Figure 4: Experimental structure.
4 EXPERIMENTAL EXAMPLE
In order to illustrate the feasibility and effectiveness
of DDPM, we construct an experimental example in
which we simulate common ICMP flooding and
SYN flooding as two types of DDoS attacks. ICMP
flooding is a type of attack that consumes the
bandwidth of victim’s network, while SYN flooding
is aiming to exhaust the resources of victim. The
experimental structure is shown as Figure 4.
In Figure 4, SDN network consists of OFSwitch
1~4, SDN controller and server. Firstly, OFSwitch 1
is used to connect with outer network 1~3. Secondly,
OFSwitch 2 connects with OFSwitch 1 by a 15M
link. Thirdly, OFSwitch 3 and OFSwitch 4 are both
linked between OFSwitch 1 and OFSwitch 2 and the
bandwidth of the links are all set for 30M. Finally,
SDN controller controls the underlying network by
southern interface. In addition, the link between
OFSwitch 1 and OFSwitch 2 is the shortest path but
not always the optimal path, which is determined by
the actual loads of each link. Data flows from
network 1 act as the normal data request, while data
flows from network 2~3 act as the abnormal ones.
Detail steps are given:
Step1: Outer network sends data flows to server.
(Network 1 sends few steady data flows, network 2
sends quantities of TCP requests accompanied with
few ICMP requests and network 3 sends quantities
of ICMP requests with few TCP requests.)
Step2: All varieties of data flows stream into
SDN network. If OFSwitch 1 cannot process the
data flows, it will deliver them to SDN controller by
Packet_In messages. And then turn to Step3.
Otherwise, turn to Step6.
Step3: SDN controller delivers the messages to
TI to analyse them. The results of analysis are
submitted to Service Layer by SDN controller.
Step4: Each module of Service Layer works
coordinately. CS develops the newest strategy for
the data flows and issues to SDN controller.
Step5: SDN controller issues the flow tables and
OFSwitch’s status tables to the relative OFSwitch
according to the strategy.
Step6: Data flows match the flow tables and
OFSwitchs set their status. Data flows’ action will
be specified to be forwarded, limited or dropped
according to the records. Then the information of
network will be updated. Turn to Step1.
The experimental process is shown as follows:
As can be seen from Figure 5, data from network
2~3 increase dramatically in different time points
respectively, while data from network 1 stay steady
throughout the experiment. When data sending rate
is over the first threshold (20Mbits/ 3s, in Figure 5
and Figure 6), the system chooses to drop the
packets with the type that leads to exceed the
threshold (at 30s from network 2 and 69s from
network 3, in Figure 6). When data sending rate is
over the second threshold (30Mbits/ 3s, in Figure 5
and Figure 6), the system drops the packets with
source ip from certain network (at 63s from network
2 and 84s from network 3, in Figure 6). Moreover, at
the beginning, the data flows are forwarding on the
link: s1 to s2. With the increasing loads of the link,
the system splits the flows (data from network 1
transmitted on link: s1 to s2; data from network 2
transmitted on link: s1 to s3; data from network 3
transmitted on link: s1 to s4) when the loads of each
link are over its threshold (s1 to s2: 10Mbits, s1 to
s3: 20Mbits, s1 to s4: 20Mbits, in Figure 7). It can
be seen that DDPM is feasible and efficient.
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture
45
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture
45
Figure 5: Data sent by outer network.
Figure 6: Data received by server.
Figure 7: Loads of each link.
5 CONCLUSIONS
DDPM achieves a new dynamic protection
mechanism to prevent DDoS attacks flexibly
according to the intensity of attacks in WLAN. The
characteristic of service atomization makes DDPM
easy to extend new function modules to improve its
reliability. Moreover, DDPM can resist the potential
DDoS attacks before the server is under attack.
Therefore, DDPM can deal with the diverse and
complicated types of DDoS attacks efficiently. In
future work, we will focus on the design of security
protection mechanism based on multiple controllers
in large scale network.
ACKNOWLEDGEMENTS
This work was supported by the National Natural
Science Foundation of China under Grant No.
61602351, No. 61502359, No. 61602349 and No.
61303117, the Young Scientist Foundation of
Wuhan University of Science and Technology under
Grant No. 2015XG005, the Open Foundation of
Hubei Province Key Laboratory of Intelligent
Information Processing and Real-time Industrial
System under Grant No. 2016znss10B, and the
College Student Innovation Foundation of Wuhan
University of Science and Technology under Grant
No. 15ZRA096. Heng He is the corresponding
author of the article.
REFERENCES
Lee, K., Kim, J., Kwon, K., Han, Y., Kim, S., 2008.
DDoS attack detection method using cluster analysis.
Expert Systems with Applications, 34(3), pp. 1659-
1665.
Lei, B., 2013. Deciphering SDN: Core Techniques and
Practical Guide. Publishing House of Electronics
Industry. Beijing.
Park, K., 2003. Scalable DDoS protection using route-
based filtering. In Proceedings of DARPA Information
Survivability Conference and Exposition, pp. 97-97.
IEEE Computer Society: Washington, DC.
Pelechrinis, K., Iliofotou, M., Krishnamurthy, S., V., 2011.
Denial of service attacks in wireless networks: The
case of jammers. Communications Surveys &
Tutorials, 13(2), pp. 245-257. IEEE.
Ramos, E., Chae, S., Kim, M., Choi, M., 2008. The
optimistic schemes of cluster analysis and k-NN
classifier method in detecting and counteracting
learned DDoS attack. In Proceedings of New
Technologies, Mobility and Security, pp. 1-5. IEEE
Computer Society: Tangier.
Thatte, G., Mitra, U., Heidemann, J., 2011. Parametric
methods for anomaly detection in aggregate traffic.
IEEE/ACM Transactions on Networking, 19(2),
pp.512-525.
Thomas, D., Ken, G., 2014. SDN: Software Defined
Networks. People's Posts and Telecommunications
Press. Beijing.
Tupakula, U., Varadharajan, V., Vuppala, S., K., 2011.
Counteracting DDoS attacks in WLAN. In
Proceedings of the 4th International Conference on
Security of Information and Networks, pp. 119-126.
ACM.
ISME 2016 - Information Science and Management Engineering IV
46
ISME 2016 - International Conference on Information System and Management Engineering
46
Xiang, Y., Lin, Y., Lei, W., Huang, S., 2004. Detecting
DDoS attack based on network self-similarity. IEE
Proceeding on Communications, 151(3), pp. 292-295.
Zhang, Y., Wan, Z., Wu, M., 2009. An active DDoS
defence model based on packet marking. In
Proceedings of the 2nd International Workshop on
Computer Science and Engineering, pp. 435-438.
IEEE Computer Society.
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture
47
A Dynamic DDoS Protection Mechanism for WLAN based on SDS Architecture
47