You are on page 1of 7

FADM: DDoS Flooding Attack Detection and

Mitigation System in Software-Defined Networking


Dingwen Hu, Peilin Hong, and Yixin Chen

Key Laboratory of Wireless-Optical Communications, School of Information Science and Technology,


University of Science and Technology of China, HeFei, Anhui 230027 China
Email: {hdw2016@mail.ustc.edu.cn, plhong@ustc.edu.cn, cyx00@mail.ustc.edu.cn}

Abstract—Distributed Denial-of-Service (DDoS) flooding at- make it easy to detect and to react against DDoS attacks [2].
tack is one of the most serious threats to network security. On the other hand, some characteristics of SDN also make
Software-Defined Networking (SDN) has recently emerged as a itself vulnerable to DDoS attacks, such as the controller is too
new network management platform, and its centralized control
architecture brings many new opportunities for defending against centralized, the switch flow table size is limited, and the band-
network attacks. In this paper, we propose FADM, an efficient width between control and data planes is a bottleneck. How
and lightweight framework to detect and mitigate DDoS attacks to take full advantage of the SDN characteristics to defend
in SDN. Firstly, the network traffic information is collected against DDoS attacks and protect the SDN infrastructure from
through the SDN controller and sFlow agents. Then an entropy- attacks is a very challenging problem.
based method is used to measure network features, and the SVM
classifier is applied to identify network anomalies. By adopting In this paper, we propose a framework to detect and mitigate
these methods together, the timeliness and accuracy of attack DDoS flooding attacks in SDN network to tackle the challenge
detection are effectively improved. To keep the major network above. Information collection is the prerequisite for attack
functionality working, we propose an efficient attack mitigation detection. According to the characteristics of SDN, we propose
mechanism based on the white-list and traffic migration. By an information collection method combining the SDN con-
introducing the mitigation agent to the network, attack traffic can
be timely blocked while benign traffic can be forwarded as usual, troller with sFlow[3] to improve its accuracy and timeliness.
which prevents the controller resources from being exhausted and Then we use an entropy-based method to measure the changes
ensures that legitimate users can access the network normally. of network features, and use the machine learning algorithm
The experimental results show that multiple DDoS attacks can to automatically detect network anomalies. By using these
be accurately detected and effectively mitigated by FADM, which methods together, DDoS attacks can be accurately detected
enables the network to recover in a short time.
in the early stage.
When the SDN network suffers from DDoS attacks, the con-
I. I NTRODUCTION
troller resources and the bandwidth between data and control
Distributed Denial-of-Service (DDoS) flooding attack is one planes will be exhausted by attack traffic. To avoid cutting off
of the most serious threats to current network security. Attack- the communication of legitimate users, we can not simply drop
ers make use of a botnet to send massive useless packets, all attack traffic, therefore we introduce the mitigation agent in
which will rapidly exhaust the target’s resources such as the network. When a DDoS attack occurs, all suspected attack
bandwidth, memory, CPU, etc. Compared with other attacks, traffic flows are migrated to the agent through the installed
DDoS attacks have various attack methods, large-scale traffic wildcard flow rules. However, some benign traffic flows may
and distributed infected hosts, all of which make it a challenge also be migrated to the mitigation agent too, so there must be
to defend. On October 21, 2016, attackers used the Mirai a way to identify them and install extra forwarding rules. By
botnet to launch two large and complex DDoS attacks against adopting this mitigation mechanism, the network can recover
the Dyn’s Managed DNS platform from 100,000 malicious quickly and provide services for legitimate users.
endpoints, which caused major Internet platforms and services The main contributions of this work can be summarized as
to be unavailable to large swathes of users in Europe and North follows:
America [1]. How to detect and defense such a large-scale 1) We propose a real-time DDoS attack detection scheme,
DDoS attack is an urgent problem to be studied. which can effectively improve the timeliness and ac-
Software-Defined Networking (SDN) has recently emerged curacy of attack detection by adopting the Controller-
as a new network management platform. In the SDN architec- based and sFlow-based information collection meth-
ture, the control and data planes are decoupled. As a result, ods, entropy-based feature extraction method and SVM-
the network intelligence and state can be logically centralized, based classification method.
and the underlying network infrastructure is abstract from 2) We propose an efficient DDoS attack mitigation mech-
applications. SDN has the capabilities of software-based traffic anism based on the white-list and dynamic updating of
analysis, logically centralized control, global view of the forwarding rules. By introducing the mitigation agent to
network, and dynamic updating of forwarding rules, which the network, attack traffic can be timely blocked while

978-1-5090-5019-2/17/$31.00 ©2017 IEEE


benign traffic is forwarded as usual, which prevents the and timeliness of attack detection, we choose SVM as the
controller resources from being exhausted and ensures classification method.
that legitimate users can access the network normally. In the SDN architecture, the dynamic updating of forward-
3) We implement an efficient, lightweight and protocol- ing rules makes it easy to defend against DDoS attacks. [6]
independent prototype system called FADM. The experi- and [14] dropped all identified attack traffic by installing flow
mental results show that it can accurately detect multiple rules. Due to the limited size of the switch flow table, it is not
DDoS flooding attacks in the early stage and enables the possible to install dropping flow rules for each attack traffic. In
network to recover quickly. [5], a White List function that maintained a list of Destination
This paper is structured as follows: Section II discusses IP addresses/ports was implemented to avoid cutting off a
the related work and the differences with our mechanism. valuable service (i.e. ftp). However, it could not avoid DDoS
Section III briefly introduces the design principles and system attacks against the service. By contrary, we introduce a white-
architecture. The implementation details of the system are list mechanism that contains a list of Source IP addresses,
presented in Section IV and the evaluation experiments are which is dynamically generated and updated based on network
presented in Section V. Finally, Section VI concludes the paper traffic information.
and discusses future work. AVANT-GUARD [15] introduced connection migration and
actuating trigger modules to defend against network attacks.
II. R ELATED W ORK But it needed to expand the data plane, and could only defend
In recent years, researchers have proposed various methods against TCP-based attacks. FloodGuard [16] introduced a data
to defend against DDoS attacks in SDN environment. Braga plane cache between data and control planes. It fowarded all
et al. [4], proposed a lightweight method for DDoS attack de- table-miss packets to the data plane cache and restrict sending
tection based on six traffic flow features. However, it used the rate during the saturation attack. But it did not classify attack
OpenFlow (OF) protocol to gather traffic information, without traffic and benign traffic, resulting in the benign traffic could
considering the control plane overloading and bandwidth bot- not be quickly forwarded. In our system, we introduce the
tleneck, which made it only suitable for small-scale networks. mitigation agent in SDN network, which does not need to
In order to improve the scalability of native OF approaches, extend the data plane, and can timely block the attack traffic
[5] proposed a new flow statistics collection method combining and quickly forward the benign traffic.
OF with sFlow. sFlow[3] is a technology for monitoring traffic III. S YSTEM D ESIGN
by the sampling mechanisms, and has been widely used to
gather traffic information [6–8]. However, due to sFlow can not A. Design Principles
gather the information of all packets, it will affect the accuracy When designing the FADM, we consider the following
of anomaly detection. To solve this problem, we propose an objectives. Firstly, the system should be accurate and sensitive
information gathering method combining the SDN controller to DDoS flooding attacks to guarantee real-time detection. In
with sFlow. In a small-scale network, we collect information addition, the attacks should be effectively mitigated to keep
directly through the controller. While in a large-scale network, the network robust. Finally, we strive to achieve the goals
we use the sFlow-based method instead. mentioned above with the whole system to be as lightweight
How to accurately measure the DDoS attack features is very and scalable as possible.
important for attack detection. In [4], six flow-based features B. System Architecture
were employed to detect DDoS flooding attacks, while the flow
The architecture of FADM consists of two main modules
rate and flow duration were used in FlowTrApp [6] to classify
as shown in Fig. 1: 1) DDoS detection module, and 2) DDoS
a traffic flow as either attack traffic or legitimate traffic.
mitigation module.
Meanwhile, entropy was used by [5] and [9] to detect network
anomalies. Compared with other features, the entropy-based
features can measure the network changes more sensitively. SDN Controller

Therefore, we adopt them to measure the features of DDoS DDoS Mitigation

attacks. DDoS Detection Mitigation


Server
With the development of machine learning technologies,
they are widely used in DDoS attack detection. In [4, 10], Control Plane
Self Organizing Maps (SOM) was used to classify network
traffic as either normal or abnormal. The high time complexity
of SOM leads to poor real-time performance in the attack Mitigation
OpenFlow Agent
detection system. In [11], the neural network was used for Protocol

calculating the risk of a DDoS attack, but its time complexity


is also high. In [7, 8, 12], Support Vector Machine (SVM) [13] Data Plane
was used to detect DDoS in SDN environment. Compared
with other methods, SVM classifier has high classification Fig. 1: System Architecture
accuracy and little processing time. To improve the accuracy
1) DDoS Detection Module: The objective of this module each new traffic flow will send a Packet-In message to the
is to detect DDoS flooding attacks in real time. This module is controller, we can collect all traffic information of the whole
running as an application in the controller, which is responsible network directly through the controller, which generates no
for collecting traffic information, extracting network features, any additional communication.
and automatically detecting network anomalies. After extensive testing, we find that the sFlow-based method
2) DDoS Mitigation Module: The purpose of this module cannot collect all traffic information when the normal traffic
is to effectively mitigate DDoS attacks to enable the network arrives at a low rate, while our proposed Controller-based
to recover quickly. This module consists of two components: (CT-based) method can still gather enough information to
i) Mitigation Server, and ii) Mitigation Agent. The mitigation maintain a higher accuracy in this condition. On the other
server is running as an application on the controller, while the hand, when the normal traffic arrives at a high rate, the switch
mitigation agent is running on a host in the SDN network. The buffers will be exhausted by normal traffic and some attack
two components work together to migrate attack traffic timely traffic will be dropped, resulting in that the CT-based method
and forward benign traffic normally. cannot collect the information completely. But the sFlow
Collector can receive the sampling packets as normal, so the
IV. S YSTEM I MPLEMENTATION
performance of the sFlow-based method is better than the CT-
A. DDoS Detection Module based method. Therefore, in order to improve the accuracy of
The procedure of this module is divided into three phases: information collection, we choose these two methods accord-
1) information collection, 2) feature extraction, and 3) attack ing to different network environments. In Small Office/Home
detection. Office (SOHO) environments, the normal traffic rate is low,
1) Information collection: Information collection is the first the CT-based method is used to collect information. While in
phase of DDoS attacks detection. In SDN environment, there larger enterprise or ISP environments, we use the sFlow-based
are two commonly used methods for collecting information. method instead.
The first one is based on OF protocol. In this method, 2) Feature extraction: Feature extraction is the prerequisite
the controller periodically sends a state request message to for DDoS attacks detection, and its purpose is to extract the
switches, and the switches respond with one or more reply current network features from the collected information. When
messages containing flow statistics. This method can gather a DDoS attack is launched, a large number of distributed
the overall traffic passing through the switch in full detail, malicious endpoints simultaneously attack a specific target,
and has been applied to collect traffic flow information in which causes the distribution of IP addresses and ports within
[4]. However, when the network suffers a high-rate DDoS the network to change. The entropy is a measure of the uncer-
attack, the bandwidth between the controller and switches is tainty of random variables in information theory. High entropy
exhausted by attack traffic. Meanwhile, the switch flow table values signify a more dispersed probability distribution, while
is full and the sizes of reply messages are very large, which low entropy values denote the concentration of a distribution.
make the channel much more congested. Finally, the controller Therefore, we use the entropy to measure the changes of
can not receive the reply messages timely, or the connection network features.
between the switch is broken. Therefore, the OF-based method We define a flow by a five-tuple {sip, dip, sport, dport,
is not suitable for detecting high-rate DDoS attacks. proto}. Assume that the total number of flows in the 𝑘 time
The second method is based on flow monitoring mechanism period is 𝑆𝑘 , the set of source IP addresses is {(𝑠𝑖𝑝𝑘𝑖 , 𝑎𝑘𝑖 )∣𝑖 =
utilizing packet samples. sFlow-RT[17] is a traffic monitoring 1, 2, ..., 𝑁 }, in which 𝑎𝑘𝑖 represents the number of flows with
software that can be used in SDN networks. In sFlow-RT, the source IP address 𝑠𝑖𝑝𝑘𝑖 in the 𝑘 time period. According to
the flow samples and counter samples are gathered by sFlow the definition of entropy, the source IP address entropy can be
Agents embedded in network devices and are sent to sFlow defined as:
Collector periodically. This method does not consume the ∑𝑁 ∑𝑁
𝑎𝑘𝑖 𝑎𝑘
bandwidth between the controller and switches, so it can 𝐻(𝑠𝑖𝑝)𝑘 = − 𝑘 𝑘
𝑝(𝑠𝑖𝑝𝑖 ) log2 𝑝(𝑠𝑖𝑝𝑖 ) = − log2 𝑖
𝑖=1
𝑆
𝑖=1 𝑘
𝑆𝑘
overcome the hindrances presented in the first method. This
method has been successfully used to collect flow statistics Similarly, we can also define the destination IP address
in [5–8]. However, due to sFlow only periodically sampling entropy 𝐻(𝑑𝑖𝑝)𝑘 , source port entropy 𝐻(𝑠𝑝𝑜𝑟𝑡)𝑘 , and des-
packets, it can not gather the information of all packets, which tination port entropy 𝐻(𝑑𝑝𝑜𝑟𝑡)𝑘 . Finally, we represent the
will cause some impact on the accuracy of anomaly detection. network features of the 𝑘 time period with a vector 𝑋𝑘 =
To address the problems of the two methods mentioned {𝐻(𝑠𝑖𝑝)𝑘 , 𝐻(𝑑𝑖𝑝)𝑘 , 𝐻(𝑠𝑝𝑜𝑟𝑡)𝑘 , 𝐻(𝑑𝑝𝑜𝑟𝑡)𝑘 }.
above, we propose a method to collect information directly For instance, when the network suffers from a DDoS
through the SDN controller. According to the OF protocol, attack, the number of source IP addresses in the network
if there is no matching flow entry when a packet arrives, will increase sharply, which results in increase in the entropy
the switch will send a Packet-In message which contains the of the source IP addresses 𝐻(𝑠𝑖𝑝); While the destination
packet header to the controller for actions. Some important IP addresses are relatively concentrated, the 𝐻(𝑑𝑖𝑝) will be
traffic information can be extracted from the packet header, decreased accordingly. If the source ports and destination ports
such as network protocol, IP addresses, ports, etc. Since of attack packets are randomly generated, the 𝐻(𝑠𝑝𝑜𝑟𝑡) and
𝐻(𝑑𝑝𝑜𝑟𝑡) will also increase significantly. Therefore, the vector 1. Detect
attacks Mitigation
𝑋𝑘 can better reflect the DDoS attack characteristics. Detection
Module Server
4. Identify
3) Attack detection: The DDoS attack detection can be 2. Install 5. Install
and forward
benign
regarded as a binary classification problem in the field of mitigation forwarding
traffics
flow rule flow rule
machine learning, which classifies the current network state Attack
traffics
as normal or abnormal based on the network feature vector Mitigation
Agent
𝑋𝑘 . Benign
3. Migrate traffics
traffics OF switch
The SVM classifier is a binary classification algorithm based
on statistical learning theory and structural risk minimization Protected
Hosts
theory. When a training sample set 𝐷 is given, the goal of
SVM is to solve the optimal partition hyperplane to minimize Fig. 2: Working process of DDoS mitigation module
the sum of the distances between the two different support
vectors to the hyperplane. With a good balance between
learning ability and complexity, SVM has been widely used Next, we introduce the detailed implementation of the three
in DDoS attack detection, only needing a small number of functions.
training samples to achieve good generalization. 2) Collecting white-list: There are two types of white lists:
To meet the system requirements for accuracy and effi- one is the static white-list set by the network administrator,
ciency, we choose SVM as the classification algorithm for which can be a single IP address or IP address segment; the
detecting DDoS attacks. The network features of pervious 𝑀 other is the dynamic white-list, which is collected periodically
time periods are applied to construct the training sample set by the mitigation server according to certain rules.
𝐷 = {(𝑋𝑖 , 𝑦𝑖 )∣𝑖 = 1, 2, ..., 𝑀, 𝑦𝑖 ∈ {−1, +1}}, in which 𝑋𝑖 If the current network is in a normal state, the mitigation
is the network feature vector of the 𝑖 time period, and 𝑦𝑖 is server stores the traffic flow information collected by the
the category of network state, ‘−1’ means ‘normal’ and ‘+1’ detection module to a database, including the collection time
means ‘abnormal’. In order to improve the training speed and information. At periodic intervals, the server analyzes the
the accuracy of the model, the data set needs to be normalized stored information for updating white-list according to the
according to the Min-Max Normalization algorithm, and the following rules:
normalized range is [-1,1]. The SVM module is trained by ∙ The flow has a pair-flow;
the data set 𝐷, and then used to classify the network feature ∙ The number of flows with the same source IP address
vector 𝑋𝑘 of the 𝑘 time period. The classification result for exceeds the threshold value (𝐶);
‘+1’ indicates that the network may be suffering from DDoS ∙ The number of days that the same source IP address
attacks. By using this method, we can accurately detect DDoS appears in the database exceeds the threshold value (𝑇 ).
attacks in real time.
The source IP addresses that satisfy the above rules are
added to the white-list and sent to the mitigation agent at the
B. DDoS Mitigation Module
same time. The threshold value of 𝐶 and 𝑇 should be set
This section introduces the working process of the DDoS according to the network scale. For large-scale networks, the
mitigation module and its three main functions. threshold should be set larger, and for small-scale networks,
1) Working process: This module contains two compo- the threshold should be set relatively small.
nents: mitigation server and mitigation agent. The working 3) Migrating attack traffic: When the network suffers from
process of this module is divided into five steps as shown in DDoS attacks, the attack traffic must be migrated or dropped to
Fig. 2: i) When a DDoS attack is detected by the detection avoid exhausting the controller resources and bandwidth. Due
module, the mitigation module enters into the attack mitigation to the limited size of the switch flow table, it is not possible to
state; ii) Then the mitigation server identifies attack traffic, install migrating or dropping flow rules for each attack traffic,
and installs a migrating flow rule which has all fields wild- so we can only install a wildcard flow rule. To avoid cutting
carded except the network protocol, destination MAC and IP off the communication of legitimate users, we can not simply
addresses; iii) All traffic flows matching the wildcard flow drop all suspected attack traffic, so we introduce the mitigation
rule are migrated to the mitigation agent, including similar agent in the network. All suspected attack traffic flows are first
attack traffic and some benign traffic; iv) The benign traffic is migrated to the mitigation agent, and then further identified
identified by the mitigation agent and will be forwarded to the by it. If the traffic is a benign traffic, it is forwarded to the
mitigation server, while the other traffic is dropped directly; mitigation server to install a forwarding flow rule.
v) The mitigation server installs a forwarding flow rule for the The mitigation server and agent identify the benign traffic
benign traffic and forwards it to the target host. and attack traffic based on the collected white-list and the
As can be seen from the above process, the mitigation server protected host list. If the source IP address is in the white-list,
and agent work together to achieve two functions: attack traffic the traffic is identified as a benign traffic and is handled by
migrating and benign traffic forwarding. To identify the attack the controller as normal. If the source IP address is not in the
traffic and benign traffic, we introduce a white-list mechanism. white-list and the destination IP address is in the protected host
list, the traffic is identified as an attack traffic for the protected agent is running on a host in the SDN network, and the two
target host. Then the mitigation server generates a wildcard components establish a TCP connection through static flow
flow rule for the attack traffic. All fields in the flow rule are rules. The mitigation agent uses the pypcap [19] library to
wildcarded other than the network protocol, destination MAC capture all packets passing through the host.
address, and destination IP address. The action of the rule For simplicity, we use Mininet to simulate the SDN net-
is set to output the traffic to the mitigation agent. After the work, including three Open-vSwitch switches and three virtual
flow rule is installed on the switch, all similar attack traffic hosts. One of the hosts is used to build a Web server, the other
matching the rule will be migrated to the mitigation agent. for running the mitigation agent, and the third for sending
4) Forwarding benign traffic: Normally, the benign traffic background traffic. Both the protected host and mitigation
will be handled directly by controller. But after the migrating agent can be extended according to the configuration informa-
flow rule is installed on the switch, if the benign traffic has tion at any time. For a protected host, we can also configure
the same network protocol and destination addresses as the multiple mitigation agents to migrate different types of traffic
attack traffic, it will also match the migrating flow rule and to different agent.
be migrated to the mitigation agent. To prevent these benign The Botnet is simulated by four hosts in external network,
traffic from being dropped, it needs to be forwarded to the which uses the TFN2K tool to launch multiple types of DDoS
target host through the mitigation server and agent working flooding attacks.
together.
The mitigation agent captures all migrated packets, and B. Training SVM Model
identifies benign traffic from them according to the white-list SVM is a supervised machine learning algorithm that needs
received from the mitigation server. If the source IP address is to be trained before using. In order to make the model more
in the white-list, the traffic is identified as a benign traffic applicable, we use the Botnet to launch different types and
and will be forwarded to the mitigation server. After the scales of DDoS attacks on the Web server, and use Tcpreplay
mitigation server receives this traffic, it generates a normal tool to replay the packets captured from the gateway of our
forwarding flow rule and set the action to output to the target laboratory, with an average traffic rate of 10 Mbps and about
host. In addition, since the packet received by the mitigation 40 active hosts. Then we extract network feature vectors from
server is the first packet of the benign traffic, it also needs the collected information periodically and set the category for
to be forwarded to the target host to complete the normal each vector. Finally, all network feature vectors are normalized
communication. and used to train the SVM model. The number of training
samples is shown in Table I.
V. E VALUATION
In this section we introduce the experimental environment TABLE I: Training samples number
and evaluate the performance and overhead of FADM. Attack samples Normal samples Total samples
SYN flood 173 286 459
A. Experimental Environment UDP flood 196 194 390
In order to evaluate the performance of FADM as accurately ICMP flood 183 182 365
and objectively , we build a experiment environment as
realistic as possible, which is shown in Fig. 3.
C. DDoS Detection Effect
FADM POX Controller The Detection Rate (DR) and False Alarm rate (FA) [4]
DDoS Detection are used to measure the performance of our proposed DDoS
Mitigation Server attacks detection scheme, which are defined as:
𝑇𝑃 𝐹𝑃
Botnet 𝐷𝑅 = , 𝐹𝐴 =
𝑇𝑃 + 𝐹𝑁 𝑇𝑁 + 𝐹𝑃
Mitigation Agent Open vSwitch
In which, 𝑇 𝑃 represents the number of attack states that are
Open vSwitch
correctly identified, while 𝐹 𝑁 represents the number of attack
Normal
Web Service Open vSwitch
Network states that are identified as normal. Therefore, 𝐷𝑅 responses
Background traffic the attack detection rate. Similarly, 𝐹 𝐴 represents the rate of
Fig. 3: Experimental Environment the normal state being incorrectly identified as the attack state.
To evaluate the performance of DDoS attacks detection, we
FADM is implemented based on the POX, and all modules have launched SYN flood attacks on the Web server under
are written in Python language. The DDoS detection module different environments as shown in Table II. Tcpreplay was
is an application of POX. It extracts the network feature vector used to generate background traffic, which could set the replay
according to the collected traffic information periodically, and speed to given rate. The time interval of information collection
then uses the LIBSVM [18] library to classify the current was 5 seconds, and the sampling rate of sFlow was 1/64.
network state as normal or abnormal automatically. The mit- As shown in Fig. 4, when the background traffic rate was
igation server is an application of POX, while the mitigation 10Mbps, the accuracy of attack detection using the CT-based
TABLE II: Parameter values used in experimentations
methods are very small. This is because both of these meth-
Background traffic rate (Mbps) Attack rate (pps) ods can collect statistical features directly from the received
Exp.1 10 100-1000 packets without the need to query switches as the OF-based
Exp.2 100 1000-15000
method.
TABLE III: Average attack detection delay
method was higher than the other two methods. This is because Background traffic CT-based (s) sFlow-based (s) OF-based (s)
the CT-based method can collect all traffic flow information 10 Mbps 3.9 4.1 12.8
of the whole network, while the sFlow-based method can only 100 Mbps 5.3 4.9 -
collect the sampling packet information. When the background
traffic rate was 100Mbps, the detection performance of the
D. DDoS Mitigation Effect
sFlow-based method was better than that of the CT-based
method, and the detection rate reached 100% when the attack To evaluate the effectiveness of the DDoS attack mitigation,
rate was greater than 3000 packets per second (pps). This three different types of DDoS attacks on the Web server were
is because the switch buffers were exhausted by background launched using the Botnet, with the attack rate increasing from
traffic, which caused some attack packets dropped as well as 1000 to 30000 pps. At the same time, we used four different
no associated Packed-In messages generated. Hence, the CT- hosts in the normal network to download the specified page
based method could not collect the information completely, from the Web server every 3 seconds, and recorded the average
but the sFlow-based method could gather the flow samples download time. At the beginning of the DDoS attack, the page
as normal. The performance of the OF-based method was the download time increased obviously, but after the migrating
worst, and it could not collect information correctly when the flow rule was installed on the switch and attack packets were
background traffic rate was 100Mbps. In addition, the values migrated to the mitigation agent, the download time returned
of 𝐹 𝐴 were always 0 in two experiments, which indicated to normal immediately. We define the average page download
that no any normal state was identified as an attack state. This time at the beginning of the attack as the network recovery
also indicated that the false alarm rate of SVM classifier was delay, and its values under various types of attacks are shown
very low. in Fig. 5.

(a) Background traffic rate = 10Mbps Fig. 5: Network recovery delay

When the network was attacked by SYN flood, the average


network recovery delay was higher than that of UDP flood
and ICMP flood. This is because in the case of SYN flood, all
TCP traffic with the protected destination IP address including
benign traffic we sent to download Web pages was migrated,
which resulted in a longer delay. In the case of two other
DDoS attacks, the benign TCP traffic was handled directly by
the controller, so the network recovery delay was short.
E. System Overhead Analysis
(b) Background traffic rate = 100Mbps FADM is mainly running on POX controller, so we evaluate
its overhead by the average CPU utilization of the controller.
Fig. 4: Detection rate under different environments We launched SYN flood attacks on the Web server with the
duration of 5 minutes and the attack rate of 30000 pps, and
Meanwhile, We use the average delay between the first monitored the CPU utilization of the controller.
attack detection time and attack launched time to evaluate the Fig. 6 shows that the CPU utilization is very low under
real-time performance of attack detection. Table III shows that normal states, which indicates that the overhead imposed
the average detection delay of the CT-based and sFlow-based by FADM is minimal. But when the DDoS attack began at
the 05:00 minute, the CPU utilization significantly increased. cloud computing environments: A survey, some research issues,
After a period of time, the attack traffic was migrated to and challenges,” IEEE Communications Surveys & Tutorials,
the mitigation agent, and the CPU utilization was restored to vol. 18, no. 1, pp. 602–622, 2016.
[3] S. Panchen, P. Phaal, and N. McKee, “Inmon corporation’s
normal. The recovery times of the three methods were 12, 15, sflow: A method for monitoring traffic in switched and routed
and 99 seconds. When using the CT-based and sFlow-based networks,” 2001.
information collection methods, the network can recover in a [4] R. Braga, E. Mota, and A. Passito, “Lightweight ddos flood-
short time, which indicates that our proposed attack detection ing attack detection using nox/openflow,” in Local Computer
and mitigation mechanism is very effective. Networks (LCN), 2010 IEEE 35th Conference on, pp. 408–415,
IEEE, 2010.
[5] K. Giotis, C. Argyropoulos, G. Androulidakis, D. Kalogeras,
and V. Maglaris, “Combining openflow and sflow for an effec-
tive and scalable anomaly detection and mitigation mechanism
on sdn environments,” Computer Networks, vol. 62, pp. 122–
136, 2014.
[6] C. Buragohain and N. Medhi, “Flowtrapp: An sdn based archi-
tecture for ddos attack detection and mitigation in data centers,”
in Signal Processing and Integrated Networks (SPIN), 2016 3rd
International Conference on, pp. 519–524, IEEE, 2016.
[7] Y. Lu and M. Wang, “An easy defense mechanism against
botnet-based ddos flooding attack originated in sdn environment
using sflow,” in Proceedings of the 11th International Confer-
ence on Future Internet Technologies, pp. 14–20, ACM, 2016.
[8] P. Wang, K.-M. Chao, H.-C. Lin, W.-H. Lin, and C.-C. Lo,
Fig. 6: Average CPU utilization of POX controller “An efficient flow control approach for sdn-based network threat
detection and migration using support vector machine,” in e-
Business Engineering (ICEBE), 2016 IEEE 13th International
VI. C ONCLUSION Conference on, pp. 56–63, IEEE, 2016.
[9] A. Lakhina, M. Crovella, and C. Diot, “Mining anomalies
In this paper, we propose FADM, an efficient and using traffic feature distributions,” in ACM SIGCOMM 2005
lightweight framework to detect and mitigate DDoS flooding Conference on Applications, Technologies, Architectures, and
attacks in SDN environment. Firstly, in order to improve Protocols for Computer Communications, Philadelphia, Penn-
sylvania, Usa, August, pp. 217–228, 2005.
the accuracy of information collection, we use the CT-based [10] Y. Xu and Y. Liu, “Ddos attack detection under sdn context,”
method and sFlow-based method according to different net- in Computer Communications, IEEE INFOCOM 2016-The 35th
work environments. Then we use the entropy-based method Annual IEEE International Conference on, pp. 1–9, IEEE, 2016.
to measure the changes of network features, and use the SVM [11] I. Mihai-Gabriel and P. Victor-Valeriu, “Achieving ddos re-
classifier to identify the current network state as normal or siliency in a software defined network by intelligent risk assess-
ment based on neural networks and danger theory,” in Compu-
abnormal. To protect the network to provide normal service tational Intelligence and Informatics (CINTI), 2014 IEEE 15th
when it suffers from DDoS attacks, we propose an efficient International Symposium on, pp. 319–324, IEEE, 2014.
attack mitigation mechanism based on white-list and dynamic [12] R. Kokila, S. T. Selvi, and K. Govindarajan, “Ddos detection
updating of forwarding rules. By introducing the mitigation and analysis in sdn-based environment using support vector
agent in the network, we can migrate the attack traffic timely machine classifier,” in Advanced Computing (ICoAC), 2014
Sixth International Conference on, pp. 205–210, IEEE, 2014.
and forward the benign traffic normally. The experimental [13] C. Cortes and V. Vapnik, “Support-vector networks,” Machine
results show that multiple DDoS flooding attacks can be Learning, vol. 20, no. 3, pp. 273–297, 1995.
accurately detected and effectively mitigated by FADM, which [14] L. V. Morales, A. F. Murillo, and S. J. Rueda, “Extending the
enables the network to recover in a short time. In addition, the floodlight controller,” in Network Computing and Applications
results show that the overhead of FADM is minimal. (NCA), 2015 IEEE 14th International Symposium on, pp. 126–
133, IEEE, 2015.
For future work, we intend to investigate using the char- [15] S. Shin, V. Yegneswaran, P. Porras, and G. Gu, “Avant-guard:
acteristics of SDN and machine learning methods to detect scalable and vigilant switch flow management in software-
application layer DDoS attacks and botnets. defined networks,” in Proceedings of the 2013 ACM SIGSAC
conference on Computer & communications security, pp. 413–
ACKNOWLEDGMENT 424, ACM, 2013.
This work is partially supported by the National Natural [16] H. Wang, L. Xu, and G. Gu, “Floodguard: a dos attack pre-
vention extension in software-defined networks,” in Dependable
Science Foundation of China under Grant No. 61671420 and Systems and Networks (DSN), 2015 45th Annual IEEE/IFIP
the Fundamental Research Funds for the Central Universities. International Conference on, pp. 239–250, IEEE, 2015.
[17] InMon, “sflow-rt.” http://www.inmon.com/products/sFlow-RT.
R EFERENCES php.
[1] S. Hilton, “Dyn analysis summary of fri- [18] C. C. Chang and C. J. Lin, LIBSVM: A library for support
day october 21 attack.” http://dyn.com/blog/ vector machines. ACM, 2011.
dyn-analysis-summary-of-friday-october-21-attack/. [19] D. Song, “pypcap.” https://github.com/pynetwork/pypcap.
[2] Q. Yan, F. R. Yu, Q. Gong, and J. Li, “Software-defined net-
working (sdn) and distributed denial of service (ddos) attacks in

You might also like