Sie sind auf Seite 1von 8

The Third International Conference on Availability, Reliability and Security

Intrusion Detection with Data Correlation Relation Graph

Amin Hassanzadeh, Babak Sadeghian


Data Security Research La (DSRL),
Computer Engineering and IT Department,
Amirkabir University of Technology
{Hassanzadeh , Basadegh}@aut.ac.ir

Abstract higher-level descriptions of attacks or a more


condensed view of the security issues highlighted
Intrusion Detection Systems are designed based on during the analysis without losing security-relevant
the assumption that the behavior of an intruder is information[2] [3].
different from a normal user of a system. We show that Alarm correlation based IDSs only determine1 the
intrusion detection can be done based on the relation and correlation between alarms, generated by
assumption that the correlation of system events and IDS’ sensors, but there are some other systems that
parameters is changed during an attack to the system. focus on alert correlation. Some of these systems,
In this paper, we propose a new method in correlating introduced in [4], do not use independent IDSs. For
data and events for "Network Based Intrusion example Ning et al. in [5] and [6] construct attack
Detection Systems". When an attack occurs, the scenarios by correlating alerts in an individual IDS on
correlation of security parameters is changed. We the basis of prerequisites and consequences of
propose to use the state of correlation between intrusions. In [1] some correlated alert create a new
parameters to detect an attack. First we show how to Meta-Alert to achieve higher-level descriptions of
select effective security parameters for our detection attacks. In [5], Ning suggests that several alert
engine with statistical correlation methods. Then, we correlation methods have been proposed for intrusion
propose how to build Correlation Relation Graphs detection systems and categorized them into three
(CRG) for the parameters showing higher correlation. classes. In the first class, the alerts are correlated based
Finally we show how the attack may be detected with on the similarities between alert attributes (for example
comparing the CRG parameter pairs for each session [7]). In the second class, the alert correlations are based
with the deviation from the regression line of them. We on scenarios which are specified by human users or
present our results for detecting a SynFlood attack learned through training datasets, and are restricted to
with this method. We give also the corresponding known attack scenarios. In the third class, the alert
detection rate and false alarm rate. correlations are the third class is based on the
preconditions and consequences of individual attacks
1. Introduction (for example see [5] and [6]).
In this paper, our approach is data correlation. Data
Evidence of attacks against a network and its correlation means associating sets of events recognized
resources can be scattered over several hosts. Intrusion through various means and applying knowledge to
detection systems have to collect and relate alert determine whether they are related, and if so, in what
information from different sources to spot complete manner and to what degree. As the amount of
attack scenarios. The process of collecting and relating correlation between two parameters is increased, the
alert information is called alert correlation. Recently, relation between parameters and their behavioral
alert correlation gained momentum and a number of similarity is more justifiable. In [8], it is suggested that
academic and commercial correlation approaches have this kind of correlation requires comparing
been suggested. However, there is no consensus on observations on different parameters, such as source
what this process is or how it should be implemented and destination IP address, an identifiable network
or evaluated [1]. Some systems use different IDSs and
then correlate the analysis results and the
corresponding alarms. This approach aims at achieving 1 This research is partially supported by Iran Telecommunication
Research Center.

0-7695-3102-4/08 $25.00 © 2008 IEEE 982


DOI 10.1109/ARES.2008.119
route, commands entered by a suspected attacker, and handshake (see Figure 1). The first step in the process
the time when activity began or ended. is a SYN packet that is sent from Source to
Each behavior is constructed with some parameters. Destination. The second message, from Destination to
For example in every TCP sessions the "protocol" field Source, has both the SYN and ACK flags set indicating
of IP packet is equal to "6", or for another example, all that Destination acknowledges the SYN and is
segments of each message in IP layer get a same value continuing the handshake. The third message, from
for identification field. In network attacks we have Source to Destination has its ACK bit set, and is an
same situation. In a TCP Flood attack, we have a lot of indication to Destination that both hosts agree that a
TCP packets with "SYN=1" and other specific connection has been established. The third message
features. But an ICMP Flood attack never has these may contain user payload data.
specifications. In fact, each attack has its own
properties that are different from other attack
properties and even with normal behavior. We use this
feature to compare each session with normal or a
specific behavior.
In this paper, we calculate the correlation between
parameters in the attack traffic with statistical methods.
If this value is greater than a defined threshold value,
the correlated parameter pairs are considered to be
included in a correlation relation graph. This threshold
is defined by an expert. We consider these final
parameters to present such an attack. This method
reduces processing load of detection engine and the set Figure 1. Three-way Handshake
of parameters that are needed for intrusion detection.
Our approach is a misuse detection approach. We also Synflood is a DoS attack to which every TCP/IP
show how to detect a Synflood attack, which is one implementation to some degree is vulnerable. Each
type of DoS attacks. We select correlated parameters of half-open TCP connection made to a machine cause
Synflood attack and calculate their regression relation. the ‘tcpd’ server to add a record to the data structure
Finally, we compare the regression relations of that stores information describing all pending
correlated parameters of each network session with connections (see Figure 2). This data structure is of
regression relations of our graph. Our approach is finite size, and it can be made to overflow by
misuse detection and we calculate the correlation of intentionally creating too many partially-open
parameters based on new observation of attack connections. The half-open connections data structure
sessions. In [9], we have shown that our method can be on the victim server system will eventually fill and the
used for anomaly detection as well. system will be unable to accept any new incoming
This paper is organized as follows. In Section 2 the connections until the table is emptied out.
structure of Neptune (Synflood) attack will be
described. In Section 3 we will describe data
correlation methods. In Section 4 the construction
algorithm of correlation relation graph will be given
and in Section 5 we will compare behavior of some
parameters in attack traffic with their behavior in
normal traffic. In Section 6 our data correlation based
detection engine and its performance for Synflood
attack detection will be presented. Finally we will give
some conclusion for the results obtained in this paper
in Section 7.

2. Synflood Attack Figure 2. Attacking to Victim with half-open


connections
Before data can be transmitted between a source
host and a destination host, TCP needs to establish a Normally there is a time out associated with a
connection between them. The connection pending connection, so the half-open connections will
establishment process is called the three-way eventually expire and the victim server system will
recover. However, the attacking system can simply

983
continue sending IP-spoofed packets requesting new security parameters with statistical analysis of an attack
connections faster than the victim system can expire behavior. In fact, we combine both statistical and rule-
the pending connections. Ref [10] believes that typical based methods. This method of data correlation
Synflood attacks can vary several parameters: the processes the gathered statistical samples of security
number of SYN packets per source address sent in a parameters from attack sessions during train phase. In
batch, the delay between successive batches, and the addition to pre-existing knowledge, this method uses
mode of source address allocation. Ref [11] says that in statistical correlation to obtain some correlation values
some cases, the system may exhaust memory, crash, or for security parameters in attack behavior.
be rendered otherwise inoperative. We call this method as Statistical Rule-based
Correlation (SRC), which calculates correlation value
3. Data Correlation Methods between parameters in attack sessions using statistical
methods. A security parameter may vary in different
All the devices, whether aimed at prevention or observations. We can find correlation value between
detection, generate huge volumes of audit data. two parameters with different pairs of observation.
Firewalls and other devices logging network Calculation of correlation value between such
connection information are especially guilty of parameters gives us the magnitude of relation between
producing vast oceans of data. Many diverse data them.
formats and representations are used for those log files In random samples of statistical population, n
and audit trails. Also, a percentage of events generated observation of X and Y variables are represented by
by network IDS and IPS are false alarms and do not ( X i , Yi ) pairs, for i=1, 2,… n. These pairs have equal
map to real threats. Further confusing issue is that the bi-variable distribution and different pairs are
different devices might report on the same things independent of each other. A simple relation between
happening on the network, but in a different way, with X and Y creates some points around the straight
no apparent way of figureuring the truth of their regression line. We use Pearson correlation coefficient
relationship. There is a definite need for a consistent to determine the value of correlation between two
analysis framework to identify various threats, parameters:
prioritize them and learn their impact on the target n

system. ∑ (X
i =1
i − X )(Yi − Y )
(1)
Correlation is defined as relationships between r=
entities, however, a good Intrusion Detection-specific  n
2
n
2
definition is lacking. From security point of view, ∑ ( X i − X )  ∑ (Yi − Y ) 
 i =1   i =1 
event correlation may be defined to improve the threat
identification and the assessment process by looking
not only at individual events, but also at their sets.
In which a ( X i , Yi ) pair is an observations of X and
Chuvakin in [12] asserts that security-specific Y random variables. X ,Y are mean values for X and Y
correlation can be loosely categorized as rule-based or respectively. r has a value in [-1, 1]. While r=1 means
statistical (algorithmic). that all points of ( X i , Yi ) pairs are on a straight line with
A rule-based correlation engine has some pre-
a positive slope. r=-1 means that all points of
existing knowledge of the attack (the rule), and it is
able to define what is actually detected in precise
( X i , Yi ) pairs are on a straight line with a negative
terms, base on that. Such attack knowledge is used to slope. When r approaches from these two values to
relate events and analyze them together in a common zero, the degree of correlation decreases; such that
context. Statistical correlation does not employ any there is no correlation in zero point.
pre-existing knowledge of the malicious activity, but
instead relies upon the knowledge (and recognition) of 3.1. Correlation Hypothesis Test
normal activities, which has been accumulated over
time. Ongoing events are then rated by a built-in The required solution for a problem might be
algorithm and may also be compared to the simplified into a choice between two competing
accumulated activity patterns, to distinguish normal hypothesis between which we have a choice, i.e., the
from abnormal (suspicious) behavior. This distinction null hypothesis, denoted by H0, against the alternative
among correlation types is somewhat similar to hypothesis, denoted by H1. The null hypothesis, H0, is
signature vs. anomaly IDS. a statement of a theory that has been put forward,
In this paper, we propose another data correlation either because it is believed to be true or because it is
method that calculates the correlation value between to be used as a basis for argument, but has not been

984
proved. The alternative hypothesis, H1, is a statement correlation coefficient between two parameters that are
of what a statistical hypothesis test is set up to calculated using (1), and its validity is examined by the
establish. hypothesis test. For example, CRG i× j represents the
We considered H0 as the null hypothesis about the
correlation coefficient r between all samples. Note that correlation value between i and j that the former
r is a random variable. We considered our H0 and H1 parameter is in the ith row and the latter one is in the
as: jth column.
We selected the entries that are greater than our
H 0 : ρ = ρ0 = 0 (2) defined threshold. Since the correlation coefficient is a
H1 : ρ ≠ ρ0(3) value in [-1, 1] and 0 ≤ r ≤ 1 , we intuitively
Where ρ is correlation coefficient of entire considered 0.5 as the appropriate threshold. It should
population and calculated by: be noted that the sign of r only shows the correlation
Cov ( X , Y ) direction.
ρ = Corr ( X , Y ) = (4)
σ XσY 4.2. CRG Construction Algorithm
The Rejection of the null hypothesis, suggests that
the alternative hypothesis may be true and r is valid. CRG is a graph for modeling a set of parameters
We can use a Confidence Value (CV) for rejecting H0. that make an Equivalence Class under the correlation
If P-value2 < (1 - CV), H0 is rejected with the relation. In fact, this graph is just a way to illustrate the
confidence value of CV. correlated parameters and analyzing the correlation
relations. There are some other ways to select
4. Correlation Relation Graph (CRG) correlated parameters such as factor analysis and
statistical clustering methods. CRG just helps the
4.1. Correlation Relation analyzer to get a graphical view of correlated
parameters and their relations. In this section we
Selecting the optimum security parameters from explain CRG construction algorithm to select
several available parameters is one of the major correlated parameters of each behavior. The related
problems of intrusion detection systems. An expert that parameters have three properties in an equivalence set,
selects these parameters for intrusion detection engine i.e.:
and gives them to the system analyzer faces some • Reflexive
challenges. One of them is determining the parameters • Symmetric
which present more effective statistical information to • Transitive
a system analyzer. The CRG construction algorithm is defined as
We examined security parameters that are follows:
correlated such that the value of their correlation 1. Each member is a CRG.
coefficient is more than threshold level introduced by 2. To add a new member to CRG two bellow
an expert. Hence, the parameters that haven’t this conditions should be satisfied:
condition are not considered for examination. a) Correlation coefficient of new member
Ref [14] introduces a set of DoS attack traffic that is with all of the previous equivalence
used for Intrusion Detection Evaluation. We select set members should be greater than the
only ten percent of all Neptune sessions of KDD threshold value.
Cup99 dataset for our statistical analysis. Each session b) The null hypothesis ( H 0 : ρ = 0 )
of KDD Cup99 dataset, either attack or normal, affects should be rejected for all of the
on 41 introduced security parameters. We calculated correlation coefficient of CRG.
the correlation coefficient between these parameters, When a new parameter is added to an equivalence
and finally introduced some of them as optimum class, it satisfies all of the above mentioned properties.
parameters for intrusion detection. Note that, each variable is correlated with itself; hence
Parameters that have required correlation value and
all of the security parameters satisfy the reflexive
participate in CRG are effective security parameters in
property. It is clear by the definition that correlation
our intrusion detection system. The first step for
relation is a symmetric relation between two
constructing CRG is to create correlation matrix of
parameters. It also remains to show that CRG has the
security parameters. Each entry of this matrix is the
third property to be an equivalence class. As each new
member should be correlated with all of the previous
2 P-value is a statistic. For more details See [13].

985
members of an equivalence class, the third property is
also satisfied.
Note that there is no necessity for the “b” condition
in values that are greater than threshold, as H0
hypothesis is rejected for high values. In practice, we
used this condition for members that have suitable
correlation with the other parameters, except one of
them. In this case if H0 was rejected but there was a
justification from an expert analyzer’s point of view, it
could be added to CRG. If a new member has a weak
correlation at least with one of CRG members, we
can’t join it to CRG because of violating equivalence
class.

4.3. CRG for Neptune (Synflood) Attack

As an example, we analyzed Neptune sessions of


KDD dataset and calculated the correlation matrix for Figure 3. Correlation Relation Graph for
its 41 parameters. Each parameter is known by an Neptune Attack
index that simplifies our demonstration of the analysis.
Nodes of CRG are the indices of security parameters
defined in KDD and the correlation value of two 5. Behavior of Neptune Detection
parameters has shown beside the corresponding edge Parameters
between them. If a CRG has more nodes it will be
more useful for Intrusion Detection Engine, as it We observe that only eleven security parameters out of
checks further parameters and increases the percentage 41 parameters are needed for detection of Neptune
of detection. For Neptune attack 11 final nodes in two attack. Behavior of these parameters changes in the
independent CRG graphs are formed (see Figure 3). two sessions of attack and normal. Hence, the
Table 1 gives the correlated parameters and correlation between the parameters for normal traffic
corresponding indices for Neptune attack sessions. will be different from those for attack traffic. Let’s call
Note that all of these parameters can be categorized in these 11 parameters as Neptune detection parameters.
3 groups, i.e., parameters of Reject Errors, parameters We show that these 11 parameters create other graphs
of Syn Errors and parameters of Number of in normal sessions and their correlation matrix is
Connections with their type of services. Parameters of different.
each group have positive correlation, while parameters We analyzed the behavior of these parameters in
in the first group (parameters of Reject Errors) and the normal traffic and calculate the correlation coefficient
second group (parameters of Syn Error) have also between them. Table 2 shows the decrease of
negative correlations. As these types of attacks correlation values. A Comparison between these values
increase Syn packets, and all of half-open connections and former values shows that in normal sessions
would be rejected, then the obtained correlation is members of second group have weak correlations and
justified. the parameter number 24 leaves its equivalence set.
These 11 parameters do not show previous correlation
Table 1. Final Parameters of Neptune Attack relations and their equivalence sets is different. Fig 4
demonstrates this difference. The correlation between
24 Srv_Count parameters number 33 and 34 decreased a few. The
25 Serror_Rate decrease is too large for the 8-node equivalence set.
26 Srv_Serror_Rate The number of its nodes is reduced to 4 nodes.
27 Rerror_Rate Parameters number 25, 26, 38 and 39 are removed
28 Srv_Rerror_Rate from the equivalence relation, and other parameters
33 Dst_Host_Srv_Count
also show less correlation.
34 Dst_Host_Same_Srv_Rate
38 Dst_Host_Serror_Rate
39 Dst_Host_Srv_Serror_Rate
40 Dst_Host_Rerror_Rate
41 Dst_Host_Srv_Rerror_Rate

986
another CRG there are 28 regression equation and
confidence interval.
We examine KDD sessions to evaluate the
operation of this method. We used 107201 sessions for
statistical analysis that resulted to two CRG such as
CRGs represented in previous sections. First, we
examined 494021 KDD sessions (10 percent of all)
with only 3-node CRG. Then these sessions evaluated
Figure 4. CRG of 11 Parameters in Normal with 8-node CRG and we used both CRGs to detect
Sessions SynFlood attacks finally. Table III shows the results of
this approach on these 494021 sessions.
We also examine completed KDD file that it has
Table. 2. Correlation Matrix of Parameters in 4898431 sessions. Table IV shows the results of
Normal Traffic evaluating all of sessions with three sets: 3N.CRG,
8N.CRG and both of them.

Table 3. Detection Results for Test on 494021


Sessions

Detected Percent False Percent


Sessions Alarm
3N.CRG 85231 79.5 34961 32.61
8N.CRG 106856 99.67 234 0.21
Both CRGs 104323 97.31 119 0.11
6. Data Correlation Based Detection
Engine
Table 4. Detection Results for Test on
In the past sections, we described our method to 4898431 Sessions
select efficient parameters that were more useful for
intrusion detection in our designed system. We
selected only 11 parameters out of 41 security Detected Percent False Percent
parameters. Now we’ll introduce our detection engine, Sessions Alarm
and the results of SynFlood attack detection will be 3N.CRG 798351 74.47 496811 46.34
8N.CRG 1060056 98.88 5953 0.55
represented.
Both CRGs 1023825 95.5 4176 0.38
Neptune detection parameters created two CRG
graphs such that one of them had 3 nodes (parameters)
and the other had 8 nodes. Our detection system looks 7. Conclusion
at correlated parameters. A detection engine that
employs these parameters should be based on In this paper, we introduced a new method for
correlation relation and the deviation from them. In the intrusion detection. For the detection of intrusions, any
section III, we pointed out the effect of correlation IDS requires some data related to the security
relation on regression line of two statistical parameters. parameters that report the security conditions of the
As the correlation value between two parameters monitored system. It is clear that increasing the number
increases, scattering of points around the regression of parameters may be useful in increasing the system
line of them will be decreased. For correlation values analyzer’s information about the monitored system,
near 1 or -1, we can say that they are over regression while gathering more information increases the amount
line completely. of information and their analysis. Some of these
We also consider a confidence interval for each security parameters are more useful than others and
regression line. This interval determines the allowed some of them are not. The analyzer should be able to
deviation for each regression equation and every pair select effective parameters from the ineffective and
that has greater intercept couldn’t justify this relation. decrease the load of processing. One of the methods
There are three regression equation and confidence for this selection is data correlation.
interval for three nodes CRG that determine the A recent trend in intrusion detection is using
relation between the parameters 24, 33, and 34. For different intrusion detection systems, and then
correlating the analysis results and the corresponding

987
alerts [2][3]. These systems only determine the relation can be performed every day or week (depending on
and correlation between alerts generated by IDS’s attacks and their modification rates). The second stage
sensors, and need to be monitored by an expert. In requires a short time to compare each event and its
these systems a new method of attack can not be parameters with regression relations. This is the main
detected because it depends on alerts generated by operation in detection engine. In fact, the former
IDS’s. We introduced a method that depends on operation that requires a long time operation never
statistical reports and any change in attack method is decreases the detection speed, as the analyzer can run it
detected. We used correlation coefficient to select in an off-line mode. If the CRGs of a behavior changed
effective parameters that have suitable correlation and in a time between two former type operations, the
can create CRGs. We intuitively considered 0.5 as the analyzer would just updates the regression relations.
appropriate threshold. For lower values we’ll have Please also notice that we have not claimed about
CRGs with more nodes and weaker links that can the speed up of this method comparing with other
increase false alarms. methods. But we have proposed an approach that helps
Then we calculated regression equation for each in reducing the number of parameters involved in
link. Each regression equation has a confidence intrusion detection and help in automating the selection
distance that formulates based on errors of each pairs of effective parameters.
with their regression equation. Finally we detected
Synflood attack sessions with calculating the deviation 8. References
of each pairs from their regression equation. Note that
if attacker changes his way of attack, these CRGs will [1] Christopher Kruegel, Fredrik Valeur, Govanni Vigna,
change and our IDS works with new CRGs. Intrusion Detection and Correlation: Challenges and
The results show that all of CRGs are not useful for Solutions, Springer (2005)
detection engine. The first CRG, that have only 3
nodes and its links are not heavy, gives more false [2] T. Chyssler, S. Nadjm-Tehrani, “Alarm Reduction and
alarms because its regression equations have greater Correlation in Defence of IP Networks”, Proceedings of the
13th International Workshops on Enabling Technologies
confidence intervals, and some normal sessions fall on (WETICE04), IEEE Computer Society, June (2004), pp. 229-
its region. The second CRG shows better results 234
because it has more nodes and they have full
correlation. If we use both CRGs, however the [3] T. Chyssler, S. Burschka, “Alarm Reduction and
detection rate decreases a little but false alarms also Correlation in Intrusion Detection Systems”, Proceedings of
decrease. Detection of Intrusions and Malware & Vulnarability
Our IDS only looks at correlated parameters and Assessment workshop (DIMVA), Gesellschaft f r Informatik,
never examine non-correlated parameters. For June (2004), pp. 9-24
example, the Synflood attacks use TCP protocol and
[4] U. Zurutuza and R. Uribeetxeberria, “Intrusion
one of the 41 parameters of KDD dataset is “protocol”. Detection Alarm Correlation: A Survey”, Proceedings of the
If this parameter existed in the CRGs formed for IADAT International Conference on Telecommunications
Synflood attack, it would be examined. If an ICMP and Computer Networks, 1-3 December, 2004
session creates one of the CRGs like these, it might
report an attack session, which increases the false [5] P.Ning, Y.Cui, and D.S.Reeves. “Constructing attack
alarms. But we could check its “protocol” parameter scenarios through correlation of intrusion alerts”, Proceeding
and ignore it. We can reduce false positives by looking of the 9th ACM Conference on Computer & Communications
at specific parameters of each alarm. Security, pages 245–254, Nov. 2002.
In fact, our IDS does not need an expert to select
[6] D. Xu, P. Ning, “Alert Correlation through Triggering
security parameters, although expert defined Events and Common Resources”, 20th Annual Computer
parameters maybe useful for detection. Thus we have Security Applications Conference, December 2004
proposed an unsupervised method to select effective
parameters. This system can work without any expert [7] Valdés, A., and Skinner, K., “Probabilistic Alert
network analyzer, so it can be modified for any new Correlation”, Proceedings of the 4th International
attack behavior. For each new unknown attack, we Symposium on Recent Advances in Intrusion detection
only need to select correlated parameters and their (RAID), Springer Verlang, California, USA, 2001, pp. 54-68
regression relation.
There are two main stages in using our data [8] Carl Endorf, Eugene Schult, Jim Mellander, Intrusion
Detection & Prevention, McGraw-Hill, (2004)
correlation based IDSs. The first stage requires relative
long time to analyze a behavior to find useful
parameters and their correlation relations. This stage

988
[9] Amin Hassanzadeh, Babak Sadeghian, “Anomaly
Intrusion Detection with Data Correlation Relation Graph”, [12] Anton Chuvakin, “Event Correlation in Security”,
Proceeding of the 11th International CSI Computer whitepaper, www.securitydocs.com, May (2004)
Conference (CSICC' 2006), Tehran, Iran, Jan. 24-26, 2006
[In Persian] [13] T. T.Soong, Fundamentals of Probability and Statistics
[10] Christopher L. Schuba, Ivan V. Krsul,et al., “Analysis of for Engineers, John Wiley and Sons Ltd, (2004)
a denial of service attack on TCP”, IEEE Computer Society
Press, May (2000) [14] Http: // kdd.ics.uci.edu / databases / kddcup99 /
task.html
[11] J.W.Haines, R.P.Lippmann, et al., “1999 DARPA
Intrusion Detection Evaluation: Design and Procedures”,
Technical report 1062, Lincoln Laboratory, MIT university,
26 February (2001)

989

Das könnte Ihnen auch gefallen