Stellar: A Fusion System for Scenario Construction and Security Risk Assessment

Size: px
Start display at page:

Download "Stellar: A Fusion System for Scenario Construction and Security Risk Assessment"

Transcription

1 Stellar: A Fusion System for Scenario Construction and Security Risk Assessment Stephen Boyer, Oliver Dain, and Robert Cunningham MIT Lincoln Laboratory Information Systems Technology Group 244 Wood St., Lexington, MA {boyer,odain,rkc}@ll.mit.edu Abstract Stellar aggregates and correlates alerts from heterogeneous network defense systems, building scenarios and estimating the security risk of the entire scenario. Prior work considered Stellar scenario formation; in this paper we explore the advantages provided by using scenario context to assess the risk of actions occurring on a network. We describe the design and an evaluation of Stellar and its Security Assessment Declarative Language (SADL), a fast, stateful, simple-to-use language for assessing the priority of scenarios, on a high traffic network under constant attack. The evaluation of the Stellar system deployed on a large, operational enterprise network demonstrated its ability to scale to high alert volumes while accurately forming and prioritizing scenarios. Stellar not only produced high priority scenarios matching all incidents reported by human analysts, but also discovered additional scenarios of concern that had initially gone unnoticed. Furthermore, by following the simple formalism embedded in example SADL rules, system administrators quickly develop a correct understanding of the network they are protecting 1. Keywords: security, fusion, scenario, correlation, intrusion detection 1 Introduction Many security-conscious organizations protect their networks with multiple computer network defense (CND) systems such as intrusion detection systems (IDSs), firewalls, 1 This work is sponsored by the Department of Defense under the Air Force Contract F C Opinions, interpretations, conclusions and recommendations are those of the author and are not necessarily endorsed by the United States Government. and routers with access control lists. While this defensein-depth strategy may improve attack detection rates, it also increases the workload of security analysts. Analysts now must expend more time and employ other resources to monitor and appropriately act on information produced by these sensors. Inundated by the large alert volume, some analysts cope by ignoring alerts that describe communication to known patched services or tune sensors to suppress such alerts. The remaining alert volume is often still large, making cross-sensor alert correlation difficult. Attack detection becomes even more difficult when network services are dynamic, multiple organizations share networks, or when users are allowed to enable and disable services on their machines. Fusion systems attempt to address these problems by providing tools to enhance the presentation and management of sensor data. In this paper we present Stellar; a real-time system that combines the alerts produced by multiple CND systems into prioritized scenarios. Stellar consists of two distinct components: the scenario building engine and the security risk assessment engine. The scenario building engine combines alerts that share a common cause. At this stage no effort is made to differentiate real attacks from false alarms. The purpose of grouping alerts is to provide a context for making security risk assessments. Security risk is determined by a set of rules written in Security Assessment Declarative Language (SADL), an easy to learn declarative language similar to the Structured Query Language (SQL) [1]. SADL is able to combine knowledge of the network environment (critical servers, mission, known vulnerabilities, etc.) and the context in which an alert appears when making an assessment. The ability to exploit context in risk assessment allows Stellar to simultaneously increase accuracy and decrease false alarm rates. Stellar was recently evaluated on a large operational network. This evaluation demonstrated the system s ability 1

2 to accurately prioritize scenarios. The system reduced the amount of data analysts had to consider by several orders of magnitude without missing any of the high priority security events identified by security professionals considering the entire alert set. Furthermore, the dynamic distributed nature of the network made it nearly impossible for the analysts to be completely aware of all the services offered on the protected network despite their skill and expertise. Stellar provides a unified framework to describe the network environment. We discovered that when administrators views of that environment (as embodied in SADL statements ) were incorrect, Stellar enabled them to quickly update their view of the network. A brief overview of related work is presented in Section 2. Stellar s approach to alert fusion is summarized in Section 3. The system architecture is described in Section 4. Section 5 provides a look at the SADL language, and the results of the recent system evaluation are presented in Section 6. 2 Related Work Network security alert correlation has become an area of active research. Much of the research has focused on discovering complex, multi-step attacks from alert data. One popular approach to this problem is to utilize sets of rules or knowledge bases to determine which alerts should be correlated. SCYLLARUS [8] and AlertSTAT [23] both employ expert knowledge to determine alert correlations. SCYLLARUS relies on the maintenance of a knowledge base which contains a list of assets, security goals, possible alerts, and the impact each alert would have on security goals. This information is used to produce hypotheses and determine which hypotheses are likely enough to report. AlertSTAT gets its expert knowledge from a set of rules written in the STATL language. Each rule specifies a set of state transitions that define the steps in a multi-step attack. A distinct rule is required for each attack scenario type. A similar approach is employed by prerequisite systems. These systems maintain information about the conditions necessary for an attack step to succeed and the conditions that are produced by a successful attack step. Multistep attack scenarios can then be discovered by searching for chains of alerts such that the post-conditions of earlier attacks satisfy the pre-conditions of later attacks. Ning and Cui describe one such system which is designed for use offline [17]. The CRIM component of MIRADOR is also designed to run offline, but includes a component to decide which scenario to report when it is possible to construct multiple plausible scenarios from a set of alerts [11, 3]. The EMERALD suite of intrusion detection tools includes a probabilistic alert correlation component [?, 22]. This system treats alerts as vectors and fuses alerts if their similarity is high enough. It is necessary to maintain a database that maps each alert type produced by the component IDSs to an attack class and a matrix that specifies the similarity between the attack classes. Most of the above systems require the maintenance of a lot of expert knowledge. For example, the prerequisite systems require the maintenance of a database of alert types, the attacks to which they correspond, and the pre and post conditions of those attacks. Stellar s correlation component requires no expert knowledge encoding. The Stellar prioritization engine, based on the SADL language, does require the user to encode some expert knowledge, but every effort is made to keep this to a minimum. Furthermore, SADL separates information about the local network configuration, which most systems administrators know, from information about common alerts and attack patterns. See Section 5 for details. The above systems assume a reasonably high fidelity alert stream. False alarms can lead to the construction of scenarios which never occurred and false negatives will fail to reveal attack steps critical to the success of the correlation mechanism. To combat the false negative problem systems must hypothesize attack steps that did not produce an alert (Cuppens and Miège call this abductive correlation [11]) but this can have an adverse effect on run time and can exacerbate the false positive problem as it is now more likely that false alarm alerts will be erroneously combined with other alerts. Combating the false positive problem is critical for systems that intend to correlate the output of the current state of the art intrusion detection tools as these sensors are not very accurate and produce a lot of false positives (some researchers have suggested that as much as 99% of typical alarm traffic is false alarms [2, 13]). If network defense systems provided the kind of detailed data described by the M2D2 data model [16] it might be possible to determine if the alerts correspond to real attacks and if those attacks were successful. However, currently deployed systems generally do not provide this kind of information making complex correlation difficult. Additionally, complex correlation systems can produce false scenarios even when the alerts are completely accurate. For example, the hosts on most networks we have examined are scanned and attacked by worms multiple times per day. There are thus many scans followed by many vulnerability exploit attempts. There are now many ways to combine these alerts (n scans followed by m exploit attempts yields n m possible scenarios) and these unrelated events may be erroneously combined to yield a scenario that looks directed and sinister. Because of these difficulties Stellar does not attempt to form complex scenarios except in the case where there is overwhelming evidence to support such a hypothesis. Instead the system performs simple correlation and focuses 2

3 on fine grained prioritization that accounts for alert context. All of the systems described above focus primarily on correlation and place little emphasis on prioritization. The M-Correlator extension to the EMERALD system (which is not the same as the probabilistic component described earlier) allows analysts to write rules, an Alert Cluster Policy Specification, to combine similar alerts [20]. These rules specify how to combine multiple alerts which correspond to the same incident but they do not attempt to correlate alerts from multiple steps of an attack scenario. An extension to the Tivoli management system correlates first by finding duplicates (multiple alerts for the same event) and then consequences (one alert being caused by another) [7]. Both types of groupings are formed by consulting a set of rules. An additional level of correlation can be performed which attempts to discover more complex scenarios called situations. While these situations can discover attack scenarios, the rules are biased toward very simple scenarios. Situations require all the alerts to have the same source IP, the same destination IP, or be of the same type. This is the type of simple correlation advocated by the authors. In contrast to Stellar, both of these systems perform prioritization on each alert in isolation instead of leveraging the context correlated alerts provide. M-Correlator s prioritization algorithm accounts for mission, alert vulnerability dependencies, and network topology, but the calculation is performed on a per-alert basis. Furthermore, like the prerequisite approaches described above, the vulnerability dependency information must be provided for each alert. The Tivoli system s duplicate and consequence rules can change the severity of an alert (e.g. raising the severity if duplicates are found) but these are configured statically and do not account for network topology or mission. Ourston et. al have developed a system that performs simple correlation and complex prioritization [18, 19]. The correlation component combines all alerts between a single pair of hosts. Prioritization is performed by Hidden Markov Models (HMMs) that attempt to match the alerts to a series of attack steps. Because the Markov Model allows for attack steps that did not produce any alerts (the hidden Markov states) the system can handle the false negative problem. The false positive problem is not directly addressed. Each possible attack scenario is encoded in the HMM and different scenarios have different priorities. This prioritization mechanism is quite different from Stellar s which utilizes a set of rules written by a domain expert. Julisch and Dacier advocate data mining historical alarms to discover alarm patterns [13]. A domain expert can then convert the discovered patterns into filters which remove common false alarms. As SADL is a powerful language that allows analysts to write the kind of rules discussed by Julisch and Dacier it might be beneficial to combine the approach they advocate with the Stellar system. Morin and Debar propose using Chronicles, a language for expressing sequences of alerts for correlation in [15]. While the time complexity of Chronicles as a function of the number of rules is addressed, the complexity as a function of the number of alerts, which is more likely to be an issue, is not addressed and could be problematic as the number of alerts grow large. 3 Stellar Approach The Stellar system defines scenarios to be collections of alerts that share a common cause. Scenarios are not predefined and the cause need not be malicious behavior. A misconfiguration or a group of false alarms could form a scenario. The constructed scenario then provides a broader context in which a priority assessment can be made. The following examples illustrate types of scenarios that an analyst might encounter: User Error Scenario: A valid user mistypes his password and fails to authenticate to a remote file server. This failed remote login attempt could trigger multiple alerts from host-based as well as network-based intrusion detection systems. All alerts caused by the failed login form a scenario. Misconfiguration Scenario: A host on the network has its default gateway incorrectly configured. When the host attempts to send a packet to a remote host, the packet is accepted through the firewall and reaches the router. The router then replies to the host with an ICMP Redirect message that is then again accepted through the firewall and received by the host. This exchange could trigger alerts from network-based IDSs, the firewall, and the router. All alerts caused by the misconfiguration form a scenario. Reconnaissance and Exploit Attempt Scenario: An adversary probes the network by scanning for hosts running a vulnerable server. At a later time, the adversary launches a known exploit. The scanning and attempted exploit activity could be observed by the firewall, as well as host-based and network-based intrusion detection systems. Once again, all alerts produced by these sensors as a result of the scanning and exploit attempt form a scenario. The intent of scenario construction is to provide context so that priority might be better assessed. Grouping alerts obviously reduces the amount of data. However, data reduction is not the only benefit. Grouping affords additional context that may reduce the number of false alarms and improve detection rates. 3

4 Alerts typically describe a single network or host event as seen from one sensor. When alerts are assessed in isolation, the analyst and/or correlation system must make a priority determination based solely on the information that a single alert provides. Alerts often provide information about real events that are nonetheless benign in order to also report on the isolated case when an event is due to a malicious operation. The benign but common operation swamps the malicious but rare operation with the predictable result being that many administrators disable the alerts that both the malicious and benign activities generate. Administrators that do not adopt this practice are faced with continuously sifting through benign alerts to determine the root cause of an event. As a result, the false alarm rates are still unacceptably high and certain types of attacks can be missed as the corresponding alerts have been disabled. Assessing risk in context allows analysts to keep low accuracy rules enabled on the devices. Typically these alerts represent false alarms and Stellar will assign the scenarios containing them a low priority. However, when the context warrants more concern, the priority of the scenario will be elevated and attacks will not be overlooked. Additionally, since alerts have not been disabled on the CND devices, potentially valuable forensic information is not lost. For example, consider the User Error Scenario above. Analysts will not normally be concerned with failed login attempt alerts as such alerts generally reflect a user forgetting or mistyping a password. As a result, security analysts might choose to disable these alerts in order to reduce the volume of false alarms. However, disabling these alerts leaves the analyst blind to password guessing attacks. Using Stellar, failed login alerts are grouped into scenarios. Those scenarios containing only a few failed login attempts could be assigned low priority and would not adversely affect analyst work load. However scenarios containing many failed login attempts could be assigned a higher priority so that password guessing attacks would not go unnoticed. In general, the evolving context allows Stellar to differentiate benign login errors from serious security threats. 4 Architecture The Stellar architecture is depicted in Figure 1. The architecture was designed with scalability in mind. Each component can run on a separate machine on different network segments. The modular design distributes the computational load and facilitates deployment customization. 4.1 Sensors, Bridges and Data Storage Because Stellar is a fusion and correlation system, it relies on the quality of the alert information produced by the CND devices deployed. As nearly every CND device produces alert data in its own format, the Stellar system employs custom bridges to aggregate, normalize, and store different sensors data in a relational database. 2 Integrating a new CND device into the Stellar system requires writing a bridge to convert the sensor s output to SQL statements that can be executed on the Stellar database. Due to the ubiquity of SQL, developers can design bridges using a wide variety of tools and computer languages. Currently, the Stellar system has bridges that accept information from RealSecure Network and Server sensors [12], Snort [21], Checkpoint firewall [14], and Battlefield Intrusion Detection System [4], a host based intrusion detection system. 4.2 The Engine, Correlation, and Prioritization The Stellar Engine has two distinct tasks. The first task is to read the alert information from the database at configurable intervals and form scenarios. Scenarios are formed based on the Scenario Generation Rules (SuGaR) that can be either explicitly written or learned through training [5],[6]. Once the scenarios are formed, the Engine assigns a priority to the entire scenario based on the SADL rules specified by a analyst. 4.3 Using Stellar The Stellar Graphical User Interface (GUI) is used for both system training and scenario monitoring. Training mode facilitates interactive learning of SuGaR, helping the analyst to discover scenarios and build examples. Modifications to the SuGaR rules would allow us to form complex scenarios using combinations of different attributes and measures. In order to evaluate the SADL engine for this paper, a simple SuGaR rule that builds scenarios from alerts arising from the same source IP address was used. Although capable of much more sophisticated correlation, this simple grouping was found to provide the best performance in terms of scenario formation on many data sets [5]. Monitoring mode displays scenarios as they are formed and prioritized. Users can assign names and comments to developing scenarios as well as produce and print scenario reports. As scenarios form, analysts can sort the scenarios by priority and quickly focus their attention on those which have been assigned high priority. Multiple analysis can cooperate by running their own local copy of the GUI and observing scenarios, SuGaR, and SADL rules. 2 IDMEF [10] is a standard for CND sensor output that could eliminate the need for bridges but is not yet widely adopted. 4

5 RealSecure NT Server Sensor Stellar Analyst GUI 1 RealSecure Network IDS Snort Database Stellar Scenario and Prioritization Engine BIDS Checkpoint Stellar Analyst GUI 2 Figure 1. Stellar Architecture 5 Security Assessment Declarative Language Once scenarios are constructed a priority is assigned. It is very useful to prioritize based on context formed from the collection of alerts in the scenario, the organization s network, mission, and behavior. The decoupling of scenario formation and prioritization to incorporate such rich context is extremely powerful and unique to Stellar. For example, an analyst may want to be able to designate that a scenario with port scan alerts from an external host targeting critical web servers is high priority. However, if the port scan is coming from the system administrator s machine between 2:00 AM and 3:00 AM, the scan does not represent a threat and the scenario is assigned a low priority. SADL allows for this kind of flexibility in scenario prioritization. Other approaches to the prioritization problem were considered. Machine learning and probabilistic approaches were rejected as building a detailed representation of the network, mission, and configuration is extremely difficult using these types of models. Expert system approaches were rejected as the languages used to specify rules tend to be complex and evaluation speed is often inadequate for this application. Using attack trees to dynamically assess security risk was also rejected due to the fact that attack trees do not account for every possible attack vector, suffer from a combinatorial explosion, and do not solve the problem of known network noise such as broken routers or weekly port scans. It is also difficult to assess ones position in an attack tree when the underlying sensor data is predominately false alarms. Having considered these as well as alternative techniques and approaches discussed in Section 2, we decided to develop an easy to use declarative language. SADL was designed to be very similar to SQL, a database query language supported by many relational databases [9]. We selected a language akin to SQL because of its power, and familiarity to target users. Conceptually, each scenario is treated as a database table. Each alert in the scenario is treated as a row in the table. Properties of the alert, such as source IP address, alert name, and time, are considered columns in the table. A SADL rule can then be viewed as a SQL statement that evaluates to either true or false. SADL rules are evaluated in order. The first rule that matches assigns the priority. For example, consider the User Error Scenario described in Section 3. We wish to differentiate Windows_Access_Error alerts caused by users forgetting or mistyping their password from brute force password guessing attacks. The goal is to assign low priority to those scenarios which contain only alerts generated by users making common errors, including Windows_Access_Error alerts, while allowing the system to generate high priority alerts for password guessing attacks and attacks that are followed by more malicious attacks. To accomplish this the group USER_ERROR_ALERTS is defined that consists of all alerts about common user errors. The two rules found in Figure 2 will then assign priority one (low priority) to scenarios containing only these alerts as long as the number of Windows_Access_Error alerts is less than 10. If more than 10 such alerts appear in a scenario the first rule will no longer match and the second rule will assign the scenario a priority of eight. A SADL rule base is typically constructed by focusing on 5

6 1: contains_only(type in USER_ERROR_ALERTS) and count(type= Windows_Access_Error) < 10; 8: count(type= Windows_Access_Error) > 9; Figure 2. Windows Access Error Example lowering the priority of the most common alerts and raising the priority of the most serious attacks. The false alarm reduction rules should be written to take advantage of all the available context and should be as specific as possible. This reduces the amount of data an administrator must consider without accidentally lowering the priority of attack. The rule above lowers the priority of scenarios containing only common user error alerts and less than ten Windows_Access_Error alerts and is a good example of a rule that will deemphasize alerts while maintaining visibility of potential attacks. A better rule would add context about the network, perhaps allowing guessing only to those systems that are known to allow remote logins. If certain assets are particularly important to an organization, rules can be written to raise the priority of attacks against these assets; particularly if the assets are vulnerable to a known class of attacks (e.g. web servers will be more vulnerable to HTTP attacks). Rules can also be written to ensure that security policy is enforced. For example, in the evaluation described in Section 6, the Windows domain controllers were of particular concern. Thus rules were written to raise the priority of any type of behavior involving these hosts that deviated from the security policy. These rules written, alerts indicating that a domain controller performed a DNS lookup against a server other than those on the list of allowed DNS servers would result in a high priority scenario. 5.1 Security Assessment Declarative Language Syntax Valid SADL rules may take one of two forms: a group statement or a priority assignment statement. A priority assignment statement consists of a number between 0 and 9 followed by a colon and a scenario level expression. A group statement consists of the keyword group, a unique group identifier, and a comma separated list in parentheses. All SADL rules are terminated with a semicolon. SADL is not case sensitive. All of the rules found in Figure 3 are valid SADL rules. The first rule the is a group statement. The other two rules are priority assignment statements. Each priority assignment statement contains a scenario level expression. Scenario level expressions are applied to scenarios and evaluate to either true or false. If the scenario level expression evaluates to true then the scenario is assigned the priority associated with the scenario level expression. Group statements simply facilitate the composition of priority assignment statements and do not assign priority to scenarios. Alert level expressions apply to properties of a single alert and are always enclosed by scenario level functions or predicates. There are only four data types in SADL: number, subnet, string, and boolean. A number is a numerical value such as 1243 or A subnet is a valid IPv4 subnet. Subnets are specified using CIDR block notation with an IPv4 address followed by a slash ( / ) and a number between 0 and 32. The number after the IP address specifies the number of 1 bits in the subnet mask. For example /16 specifies a 16 bit subnet mask which is equivalent to a mask of If no subnet mask is provided the address refers to a single host. String constants are always enclosed in quotes. Examples of valid strings include Port_Scan, RealSecure, and Monday. A boolean is a truth value: either true or false. Boolean values are usually produced by evaluating SADL predicates. Alert level expressions are formed by combining alert level properties, predicates and operators into an expression that evaluates to either true or false for a given alert. Valid alert properties are listed in Table 2. Valid operators are listed in Table 1. There are two alert level predicates: time, and *. The time predicate is used to specify a time window. This is probably best explained by example: time(02:00:00,2m) evaluates to true on any alert that occurred between 02:00 and 02:02 hours. Times are always specified in military time helping avoiding any confusion between A.M. and P.M. hours. The m after the numeral 2 in the example specifies that the time scale for the window is in minutes. Valid time scales are s for seconds, m for minutes, and h for hours. Thus time(13:00:00,2h) evaluates to true on any alert that occurred between 13:00 and 15:00 hours. There is also one special alert level predicate, the * predicate. * always returns true just like the SQL * predicate. It is useful in conjunction with some of the scenario level functions. For example, count(*) returns the number of alerts in the scenario. Scenario level functions are similar to SQL domain aggregate functions. They operate on the alerts that match the alert level expressions they enclose. For example, the 6

7 group important_ips( , /24); 8: CONTAINS(from in important_ips); 2: contains_only(type= Port_Scan ); Figure 3. Example SADL rules Operator Type Valid Data Types Example = binary string, subnet, number from= > binary number count(from= )>5 < binary number rate(src_ip= )<5 >= binary number count(dest_ip= )>=5 <= binary number count(to= )<=5 not unary boolean not type= Port_Scan in unary boolean Port_Scan in scan_group Table 1. Valid Operators Property Data Type Description Example Values from, src_ip subnet source IP address to, dest_ip subnet destination IP address type string type of alert Port_Scan protocol string protocol associated with the alert UDP, ICMP src_port number source port of the alert 2765, 80 dest_port number destination port of the alert 80, 21 day string day of the week Monday, Tuesday sensor subnet IP address of the sensor Table 2. Valid Alert Level Properties Function Return Type Description contains boolean True if the scenario contains at least 1 alert matching the alert level expression contains_only boolean True if the scenario contains only alerts matching the alert level expression contains_pct number Returns the percent of alerts in the scenario that match the alert level expression count number Returns the number of alerts that match the alert level expression count_distinct number Returns the number of distinct alerts that match the alert level expression rate number Returns the number of alerts that match the alert level expression divided by the number of seconds between the first and last such alert. Table 3. Valid Scenario Level Functions 7

8 count function takes an alert level expression as an argument and returns the number of alerts in the scenario that match the given expression. Thus count(src_ip = and type= Port_Scan ) would return the number of alerts in the scenario whose source address is and whose alert type is Port_Scan. Valid Scenario level functions are listed in Table Predicate Complexity A naive implementation of scenario level functions would require a computation time that is linear in the number of alerts in the scenario. For example, the count function must examine each alert in the scenario to see if it matches the alert level expression and count the number of times an alert does in fact match. Since scenarios are evaluated against all SADL rules each time they change, a constant time solution is desired. All of the functions listed in Table 3 maintain state, enabling a scenario s priority to be evaluated in constant time, using information stored during the last evaluation of the scenario. By maintaining state, the value of the function can be determined by examining only the alerts that were added to the scenario since it was last evaluated. Since a scenario s prioritization is re-evaluated each time alerts are added, functions can generally determine their value by examining only a single alert. 6 System Evaluation In order to evaluate the performance of Stellar we deployed the system on a large, operational enterprise network. We ran Stellar in parallel with an existing system that had separate user interfaces for each sensor and required human analysts to manually correlate and prioritize alerts into incident reports. The site s analysts helped us install, run, and provided feedback on the Stellar s performance and usability. 6.1 Site Deployment The trial network is composed of two class B networks and contains over fifty different web, DNS, mail and domain controller servers in addition to thousands of desktop hosts. Daily aggregate ingress and egress traffic rates averaged approximately 13 Mbps. Defending the network were three Checkpoint firewalls and four Snort sensors located across multiple network segments. Other CND devices were deployed but were not included in this evaluation primarily due to unsupported sensor types and time synchronization problems. The organization overseeing the network is extremely security conscious and has a team of knowledgeable and dedicated security analysts monitoring alerts twenty four hours a day, seven days a week. 6.2 Deployment Architecture and Configuration In an effort to simplify the installation process and decrease the operational impact of the evaluation, the authors pre-installed the Stellar components on laptops. The machine on which the Stellar Engine ran was a Pentium III 1.3 GHz machine with 1 GB of RAM running Windows 2000 Professional. The evaluation was intended to focus on SADL performance. As prior work had indicated that forming scenarios based on source IP address alone could be effective [5], the scenario building engine was configured to combine alerts into a scenario if and only if the source IP addresses of the alerts were identical. The evaluation took place over three days. On the first day the system was configured and deployed. Some of the analysts familiar with the installation site wrote SADL rules with the help of the authors. The rules written on the first day were deployed on the second day when Stellar was run on live operational data. During this period, the team of approximately 3 local analysts performed their duties as usual, ignoring the output from Stellar. Analysts at this site determine if detected activity might compromise network security and write incident reports. On the third day, the incident reports written during the second day were compared to the scenarios given high priority by the Stellar system. If the system was effective, all security incidents identified by the analysts would have been given high priority by the Stellar system. The third day was also used to gather feedback from the analysts about the usability and usefulness of the system. The majority of the deployment and configuration effort was composing the site specific SADL rules. SADL rule composition efforts required collaboration of several analysts in order to provide information about critical network assets, services, sensors, and policies. The process of writing such a rule set was more time intensive than originally anticipated. This experience leads us to believe that systems which require additional information will be difficult to maintain unless data input can be automated. While writing the prioritization rules took only a couple of hours (including the time required to explain the SADL language and syntax to the analysts), filling in the group statements which identify network assets was time consuming. Knowledge of all the network s services did not reside in one place or with one analyst, although nearly all said they had a mental map of the network. Compiling a comprehensive list of servers required several phone calls to other, remotely located analysts and system administrators. Additionally, initial Stellar runs revealed legitimate servers that the analysts were not aware of. These servers came to light when SADL produced high priority scenarios indicating a modest number of successful connections to unknown servers. This occurred 8

9 several times before all legitimate servers were identified. While composing the SADL rules it became apparent that many of rules could be written once and applied generally at many different sites; requiring changes only to the group statements. The rule in Figure 4 is designed to identify port 80 connections to hosts which are not known to be web servers where the connection was allowed by the firewall. To use this rule at a different site, only the elements of the two group statements (for http_servers and system_networks) would need to be modified. Many similarly valuable rules can be written. Despite the potential general applicability of a few base rules, each site will pursue its unique mission and enforce different policy necessitating the composition of custom rules to achieve the best performance. Work to reduce the composition and maintenance burden is addressed in Section 6.4. Despite the time required to compose and test the SADL rule base, the language proved to be powerful and flexible, allowing the analysts to write complex rules with few limitations. The final rule base used to perform the evaluation consisted of 18 group and 26 priority assignment statements. The group statements primarily characterized the network defining lists of hosts, services, and classes of alert types. The priority statements made use of the group statements in defining complex policies and known network behavior. Approximately one half of the priority statements addressed scenarios that were considered high priority. The remainder of the priority statements lowered priority for scenarios that were indicative of false alarms or known network behavior (i.e. misconfigurations, faulty hardware, etc.). 6.3 Quantitative System Performance and Results Once an acceptable set of SADL rules were composed and the system installed, a twenty four hour uninterrupted run of Stellar was initiated on the live network. Within the twenty four hour period, the seven sensors produced a total of 845,236 alerts. From those 845,236 alerts, Stellar formed 12,137 scenarios. Of the 12,137 scenarios formed, only sixty two were assigned a priority of seven or above Stellar Speed During the run the Stellar system did not have any problems keeping up with the alert volume. Sensors produced an average of ten alerts per second during the twenty four hour period. Experiments after the evaluation showed that the Engine was able to start, retrieve all 845,236 alerts from the database, form the 12,137 scenarios, assign priorities, and update the database in less than twelve minutes, processing and average of 1,380 alerts per second. Stellar is capable of processing alerts streams over one hundred times the rate at which the 7 sensors in the evaluation produced them Stellar Memory Use and Scenario Time Out Because the Stellar system is designed to run in real time, all the scenarios and the alerts they contain must be held in memory. When the machine on which the Engine is running runs out of physical memory, the Engine must time out scenarios. Scenarios which were the last to have alerts added to them are removed from memory first. When a scenario is removed from memory it still exists in the database and may be viewed by the analyst, but the Engine will no longer prioritize or add new alerts to timed out scenarios. Since attackers may employ long latencies in order to avoid detection, the Stellar system tries to avoid timing out scenarios by minimizing the memory requirements of the stored scenario information. Experimentation revealed that an alert requires an average of 350 bytes of memory. Configurable runtime parameters allow users to set engine thresholds for scenario time outs. During the twenty four hour run we configured the Engine to maintain no more than 400,000 alerts in memory. When this threshold was exceeded, alerts were timed out in fifty thousand alert blocks. Each time the Engine reached the 400,000 alert threshold, fifty thousand alerts were removed from memory. During the run Stellar never consumed more than 150 MB of memory. We could have maintained more alerts in memory before timing out, but we wanted to test the time out mechanism and so chose this lower threshold to ensure that time outs would be triggered several times during our test. At the end of the twenty four hour period, 3,723 scenarios remained active in memory and 8,429 scenarios had been time timed out Prioritization Accuracy During the same twenty four hour period that the Stellar system ran, site analysts continued their task of network surveillance and incident reporting. Incident reports submitted by site analysts gave us a base line from which we could assess detection performance. Site analysts submitted twelve incident reports during the twenty four hour evaluation period. All twelve incident reports concerned reconnaissance or probing activity and were among the sixty two Stellar assigned a high priority (a priority seven or greater). All alerts referenced in the incident reports appeared in the corresponding high priority scenario. The number of alerts per scenario ranged from 16 to 7456 with an average of Having confirmed successful scenario construction and prioritization for all twelve incidents reported by analysts, we further examined scenarios for which no incident had been reported but had been prioritized high. This analysis identified an additional eight scenarios that site analysts confirmed would have been the subject of an incident report had they been identified. Several of these incidents 9

10 8: contains( dest_port = 80 AND not dest_ip in http_servers AND type = accept AND dest_ip in system_networks ); Figure 4. Unaccounted for web server connection SADL rule indicated a probe followed by HTTP and SMTP exploit attempts against site servers. Fortunately none of the targeted machines were vulnerable to the attacks. The formation and high prioritization of the multi-step attack scenarios is significant in that alerts combined across multiple sensors to reveal a broader vision of attack activity. The eight unreported scenarios usually incorporated alerts from both the firewalls and the Snort sensors and did not match any existing SADL rule. Unexpected activity is assigned the default priority which is set high to catch novel attack behavior. An example of a SADL rule that assigned a high priority to scenarios confirmed to have been the subject of analysts incident reports is presented in Figure 5. This rule is a conjunction of two scenario level predicates that assigned a priority seven to any scenario that contains more than 10 firewall drop alerts to distinct hosts with fewer than five distinct destination ports. This rule is designed to identify behavior that is indicative of a horizontal network scan. Since the rule does not specify a time window is can detect slow, stealthy scans which are overlooked by Snort s port scan preprocessor. Of the sixty two high priority scenarios, a total of twenty were confirmed to warrant further investigation. Not every scenario that received a high priority necessitated an incident report. The majority of the high priority scenarios were intentionally prioritized as such by the analysts in order to keep potentially suspicious activity in view where it could immediately be acted on or dismissed. Scenarios of this sort typically received a priority of 7 where scenarios of real concerned received a 8 or 9. The overwhelming majoring of the remaining scenarios prioritized high but not reported upon constituted initial indications of reconnaissance activity. There were three scenarios for which we could not assign the desired prioritization due to limitations in SADL. This limitation is due to the fact that SADL cannot account for alert ordering in prioritization rules. Work since the evaluation has addressed this limitation and is explained in Section Qualitative Results Analysts noted that it would be easier to investigate incidents if the GUI could indicate which alert in a scenario caused the priority to change. The ability to cross reference additional data, such as port to service mappings, packet dump data, and vendor alert information was also requested. The most difficult part of writing the SADL rules was compiling a list of servers and services known to be running on the protected network. Although once the system is configured to model the network, we anticipate that it will change only slowly. Sometimes analysts will want to know about these changes, and sometimes analysts will want the system to automatically learn about these changes and adapt. To address this we are working to incorporate LANScape, a passive network monitoring tool. LANScape tracks of all hosts on the network and the services they are running. If a new host or service appears on the network the administrator is notified. Newly discovered servers and services can be marked as allowed or unknown by the administrator. Thus, for example, newly discovered HTTP servers that are marked as allowed by the administrator can be automatically added to http_servers group. Active scanning could also be used to help populate and maintain such groups. Many signature based IDSs suffer from high false positive rates partly due the fact that they are not aware of all of the services running on all the variety of OSs that the IDSs are tasked to protect. The addition of LANScape will also enable SADL to assign priority based on automatically integrated additional context such as knowledge about whether a host is listening on a targeted port or running a particular version of a vulnerable operating system. Figure 6 contains an example of what a SADL rule could look like with this extension. This rule assigns high priority to scenarios that contain web attacks destined for a host which is not on the list of allowed web servers but is none the less listening on port 80. A limitation to the SADL implementation briefly mentioned in Section does not allow one to prioritize based on the the order in which alerts occur. In a few instances, analysts wanted the ability to prioritize scenarios containing the same types of alerts differently based on the order in which those events occurred. One ordering was benign where another ordering was indicative of malicious intent. In order to overcome this temporal ordering limitation we are currently developing the ability to assign attributes to events that match arbitrary length sequences. Once events match the sequence and are assigned the sequence attribute, we will be able to prioritize based on that attribute just like any other such as type or protocol. 10

11 7: count_distinct(dest_ip, type="drop") > 10 and count_distinct(dest_port, type="drop") < 5; 9: contains(type in web_attacks and not (to in web_servers) and is_listening( dest_ip, 80 )); Figure 5. Scan and Attack SADL rule Figure 6. Potential SADL rule with machine service information 7 Conclusion The Stellar system demonstrates the benefits of exploiting context in prioritization. To support this Stellar first combines related alerts into scenarios and then prioritizes the entire scenario using a declarative prioritization language. A test deployment in an operational enterprise network demonstrated Stellar s ability to reduce analyst workload without compromising security. Stellar had no difficulty processing nearly one million alerts per day. The correlation and prioritization technique discovered all the incidents identified and reported by human analysts as well as many others that would have necessitated an incident report submission had analysts been aware of scenario s existence. The massive amount of data reduction provided by Stellar allows analysts to more quickly focus on incidents of interest as well as understand the network they protect and the traffic traversing it. References [1] M. M. Astrahan and D. D. Chamberlin. Implementation of a structured english query language. Communications of the ACM, 10: , [2] Tim Belcher and Elad Yoran. Riptech internet security threat report VII. Technical report, Riptech, July [3] S. Benferhat, F. Autrel, and F. Cuppens. Enhanced correlation in an intrusion detection process. In Proceedings of the Second International Workshop Mathematical Methods, Models and Architectures for Computer Networks Security, St. Petersburg, Russia, September [4] Robert K. Cunningham, Richard P. Lippmann, David Kassay, Seth E. Webster, and Marc A. Zissman. Hostbased bottleneck verification efficiently detects novel computer attacks. In IEEE Military Communications Conference Proceedings, Atlantic City, NJ, [5] Oliver Dain and Robert K. Cunningham. Fusing A Heterogeneous Alert Stream Into Scenarios. Kluwer Academic Publishers, [6] Oliver Dain, Robert K. Cunningham, and Stephen Boyer. IREP++, a faster rule learning algorithm. In Proceedings of the Fourth SIAM International Conference on Data Mining, Orlando, FL, April [7] Hervé Debar and Andreas Wespi. Aggregation and correlation of intrusion-detection alerts. In W. Lee, L. Mé, and A. Wespi, editors, Recent Advances in Intrusion Detection, page 85, Springer, October [8] Robert P. Goldman, W.L. Heimerdinger, Steven A. Harp, Christopher Geib, Vicraj Thomas, and Robert L. Carter. Information modeling for intrusion report aggregation. In Proceedings of the DARPA Information Survivability Conference and Exposition, pages IEEE Computer Society, June [9] James R. Groff and Paul N. Weinberg. SQL The Complete Reference. McGraw Hill, [10] Intrusion Detection Working Group. Intrusion detection message exchange format. IETF-IDWG Draft, January [11] IEEE Computer Society Technical Committee on Security and Privacy. Alert Correlation in a Cooperative Intrusion Detection Framework, Los Alamitos, CA, May [12] Internet Security Systems. RealSecure console user guide. Atlanta, GA, [13] K. Julisch and M. Dacier. Mining intrusion detection alarms for actionable knowledge, [14] Check Point Firewall Technologies LTD. FireWall-1 User Guide, 3.0 edition, html. 11

12 [15] Benjamin Morin and Hervé Debar. Correlation of intrusion symptoms: an application of chronicles. In Proceedings of the 6th symposium on Recent Advances in Intrusion Detection (RAID 2003), Pittsburg, PA, September Springer LNCS. [16] Benjamin Morin, Ludovic Mé, Hervé Debar, and Mireille Ducassé. M2D2 : a formal data model for IDS alert correlation. In Proceedings of the 5th symposium on Recent Advances in Intrusion Detection (RAID 2002), pages , Zurich, Switzerland, October Springer LNCS. [17] Peng Ning and Yun Cui. An intrusion alert correlator based on prerequisites of intrusions. Technical Report TR , [18] Dirk Ourston, Sara Matzner, William Stump, and Bryan Hopkins. Applications of hidden markov models to detecting multi-stage network attacks. In Proceedings of the 36th Annual Hawaii International Conference on System Sciences, Big Island, Hawaii, January IEEE Computer Society. [19] Dirk Ourston, Sara Matzner, William Stump, and Bryan Hopkins. Coordinated internet attacks: Responding to attack complexity. Journal of Computer Security, 12(2): , [20] Phillip A. Porras, Martin W. Fong,, and Alfonso Valdes. A mission-impact-based approach to IN- FOSEC alarm correlation. In Lecture Notes in Computer Science, Proceedings Recent Advances in Intrusion Detection, pages , Zurich, Switzerland, October [21] Marty Roesch. Snort: Lightweight intrusion detection for networks. In Usenix 13th Systems Administration Conf., Usenix Assoc. [22] Alfonso Valdes and Keith Skinner. Probabilistic alert correlation. In Recent Advances in Intrusion Detection (RAID 2001), number 2212 in Lecture Notes in Computer Science. Springer-Verlag, [23] G. Vigna, F. Valeur, and R.A. Kemmerer. Designing and Implementing a Family of Intrusion Detection Systems. In Proceedings of the European Software Engineering Conference and ACM SIGSOFT Symposium on the Foundations of Software Engineering (ESEC/FSE 2003), Helsinki, Finland, September

INTRUSION DETECTION ALARM CORRELATION: A SURVEY

INTRUSION DETECTION ALARM CORRELATION: A SURVEY INTRUSION DETECTION ALARM CORRELATION: A SURVEY Urko Zurutuza, Roberto Uribeetxeberria Computer Science Department, Mondragon University Mondragon, Gipuzkoa, (Spain) {uzurutuza,ruribeetxeberria}@eps.mondragon.edu

More information

Architecture Overview

Architecture Overview Architecture Overview Design Fundamentals The networks discussed in this paper have some common design fundamentals, including segmentation into modules, which enables network traffic to be isolated and

More information

Network- vs. Host-based Intrusion Detection

Network- vs. Host-based Intrusion Detection Network- vs. Host-based Intrusion Detection A Guide to Intrusion Detection Technology 6600 Peachtree-Dunwoody Road 300 Embassy Row Atlanta, GA 30348 Tel: 678.443.6000 Toll-free: 800.776.2362 Fax: 678.443.6477

More information

Intrusion Detection Systems with Correlation Capabilities

Intrusion Detection Systems with Correlation Capabilities Intrusion Detection Systems with Correlation Capabilities Daniel Johansson danjo133@student.liu.se Pär Andersson paran213@student.liu.se Abstract Alert correlation in network intrusion detection systems

More information

Taxonomy of Intrusion Detection System

Taxonomy of Intrusion Detection System Taxonomy of Intrusion Detection System Monika Sharma, Sumit Sharma Abstract During the past years, security of computer networks has become main stream in most of everyone's lives. Nowadays as the use

More information

STANDARDISATION AND CLASSIFICATION OF ALERTS GENERATED BY INTRUSION DETECTION SYSTEMS

STANDARDISATION AND CLASSIFICATION OF ALERTS GENERATED BY INTRUSION DETECTION SYSTEMS STANDARDISATION AND CLASSIFICATION OF ALERTS GENERATED BY INTRUSION DETECTION SYSTEMS Athira A B 1 and Vinod Pathari 2 1 Department of Computer Engineering,National Institute Of Technology Calicut, India

More information

Security Event Management. February 7, 2007 (Revision 5)

Security Event Management. February 7, 2007 (Revision 5) Security Event Management February 7, 2007 (Revision 5) Table of Contents TABLE OF CONTENTS... 2 INTRODUCTION... 3 CRITICAL EVENT DETECTION... 3 LOG ANALYSIS, REPORTING AND STORAGE... 7 LOWER TOTAL COST

More information

CS 356 Lecture 17 and 18 Intrusion Detection. Spring 2013

CS 356 Lecture 17 and 18 Intrusion Detection. Spring 2013 CS 356 Lecture 17 and 18 Intrusion Detection Spring 2013 Review Chapter 1: Basic Concepts and Terminology Chapter 2: Basic Cryptographic Tools Chapter 3 User Authentication Chapter 4 Access Control Lists

More information

IDS / IPS. James E. Thiel S.W.A.T.

IDS / IPS. James E. Thiel S.W.A.T. IDS / IPS An introduction to intrusion detection and intrusion prevention systems James E. Thiel January 14, 2005 S.W.A.T. Drexel University Overview Intrusion Detection Purpose Types Detection Methods

More information

Intrusion Detections Systems

Intrusion Detections Systems Intrusion Detections Systems 2009-03-04 Secure Computer Systems Poia Samoudi Asli Davor Sutic Contents Intrusion Detections Systems... 1 Contents... 2 Abstract... 2 Introduction... 3 IDS importance...

More information

How To Protect Your Network From Intrusions From A Malicious Computer (Malware) With A Microsoft Network Security Platform)

How To Protect Your Network From Intrusions From A Malicious Computer (Malware) With A Microsoft Network Security Platform) McAfee Security: Intrusion Prevention System REV: 0.1.1 (July 2011) 1 Contents 1. McAfee Network Security Platform...3 2. McAfee Host Intrusion Prevention for Server...4 2.1 Network IPS...4 2.2 Workload

More information

Intrusion Alert Correlation Technique Analysis for Heterogeneous Log

Intrusion Alert Correlation Technique Analysis for Heterogeneous Log 132 Intrusion Correlation Analysis for Heterogeneous Log Robiah Yusof, Siti Rahayu Selamat, Shahrin Sahib Faculty of Information Technology and Communication, Universiti Teknikal Malaysia Melaka, Ayer

More information

NETWORK SECURITY (W/LAB) Course Syllabus

NETWORK SECURITY (W/LAB) Course Syllabus 6111 E. Skelly Drive P. O. Box 477200 Tulsa, OK 74147-7200 NETWORK SECURITY (W/LAB) Course Syllabus Course Number: NTWK-0008 OHLAP Credit: Yes OCAS Code: 8131 Course Length: 130 Hours Career Cluster: Information

More information

PROTECTING INFORMATION SYSTEMS WITH FIREWALLS: REVISED GUIDELINES ON FIREWALL TECHNOLOGIES AND POLICIES

PROTECTING INFORMATION SYSTEMS WITH FIREWALLS: REVISED GUIDELINES ON FIREWALL TECHNOLOGIES AND POLICIES PROTECTING INFORMATION SYSTEMS WITH FIREWALLS: REVISED GUIDELINES ON FIREWALL TECHNOLOGIES AND POLICIES Shirley Radack, Editor Computer Security Division Information Technology Laboratory National Institute

More information

Overview. Firewall Security. Perimeter Security Devices. Routers

Overview. Firewall Security. Perimeter Security Devices. Routers Overview Firewall Security Chapter 8 Perimeter Security Devices H/W vs. S/W Packet Filtering vs. Stateful Inspection Firewall Topologies Firewall Rulebases Lecturer: Pei-yih Ting 1 2 Perimeter Security

More information

INTRUSION DETECTION SYSTEMS and Network Security

INTRUSION DETECTION SYSTEMS and Network Security INTRUSION DETECTION SYSTEMS and Network Security Intrusion Detection System IDS A layered network security approach starts with : A well secured system which starts with: Up-to-date application and OS

More information

SURVEY OF INTRUSION DETECTION SYSTEM

SURVEY OF INTRUSION DETECTION SYSTEM SURVEY OF INTRUSION DETECTION SYSTEM PRAJAPATI VAIBHAVI S. SHARMA DIPIKA V. ASST. PROF. ASST. PROF. MANISH INSTITUTE OF COMPUTER STUDIES MANISH INSTITUTE OF COMPUTER STUDIES VISNAGAR VISNAGAR GUJARAT GUJARAT

More information

Evaluating Intrusion Detection Systems without Attacking your Friends: The 1998 DARPA Intrusion Detection Evaluation

Evaluating Intrusion Detection Systems without Attacking your Friends: The 1998 DARPA Intrusion Detection Evaluation Evaluating Intrusion Detection Systems without Attacking your Friends: The 1998 DARPA Intrusion Detection Evaluation R. K. Cunningham, R. P. Lippmann, D. J. Fried, S. L. Garfinkel, I. Graf, K. R. Kendall,

More information

How To Prevent Hacker Attacks With Network Behavior Analysis

How To Prevent Hacker Attacks With Network Behavior Analysis E-Guide Signature vs. anomaly-based behavior analysis News of successful network attacks has become so commonplace that they are almost no longer news. Hackers have broken into commercial sites to steal

More information

Efficient Security Alert Management System

Efficient Security Alert Management System Efficient Security Alert Management System Minoo Deljavan Anvary IT Department School of e-learning Shiraz University Shiraz, Fars, Iran Majid Ghonji Feshki Department of Computer Science Qzvin Branch,

More information

A Review of Anomaly Detection Techniques in Network Intrusion Detection System

A Review of Anomaly Detection Techniques in Network Intrusion Detection System A Review of Anomaly Detection Techniques in Network Intrusion Detection System Dr.D.V.S.S.Subrahmanyam Professor, Dept. of CSE, Sreyas Institute of Engineering & Technology, Hyderabad, India ABSTRACT:In

More information

CHAPTER 3 : INCIDENT RESPONSE FIVE KEY RECOMMENDATIONS GLOBAL THREAT INTELLIGENCE REPORT 2015 :: COPYRIGHT 2015 NTT INNOVATION INSTITUTE 1 LLC

CHAPTER 3 : INCIDENT RESPONSE FIVE KEY RECOMMENDATIONS GLOBAL THREAT INTELLIGENCE REPORT 2015 :: COPYRIGHT 2015 NTT INNOVATION INSTITUTE 1 LLC : INCIDENT RESPONSE FIVE KEY RECOMMENDATIONS 1 FIVE KEY RECOMMENDATIONS During 2014, NTT Group supported response efforts for a variety of incidents. Review of these engagements revealed some observations

More information

Network and Host-based Vulnerability Assessment

Network and Host-based Vulnerability Assessment Network and Host-based Vulnerability Assessment A guide for information systems and network security professionals 6600 Peachtree-Dunwoody Road 300 Embassy Row Atlanta, GA 30348 Tel: 678.443.6000 Toll-free:

More information

False Alert Reduction and Correlation for Attack Scenarios with Automatic Time Window

False Alert Reduction and Correlation for Attack Scenarios with Automatic Time Window False Alert Reduction and Correlation for Attack Scenarios with Automatic Time Window M. Logaprakash Department of CSE (PG) Sri Ramakrishna Engineering College Coimbatore, India Abstract - The Intrusion

More information

This chapter covers the following topics:

This chapter covers the following topics: This chapter covers the following topics: Components of SAFE Small Network Design Corporate Internet Module Campus Module Branch Versus Headend/Standalone Considerations for Small Networks C H A P T E

More information

Intrusion Detection in AlienVault

Intrusion Detection in AlienVault Complete. Simple. Affordable Copyright 2014 AlienVault. All rights reserved. AlienVault, AlienVault Unified Security Management, AlienVault USM, AlienVault Open Threat Exchange, AlienVault OTX, Open Threat

More information

Intrusion Detection Systems

Intrusion Detection Systems Intrusion Detection Systems Assessment of the operation and usefulness of informatics tools for the detection of on-going computer attacks André Matos Luís Machado Work Topics 1. Definition 2. Characteristics

More information

Second-generation (GenII) honeypots

Second-generation (GenII) honeypots Second-generation (GenII) honeypots Bojan Zdrnja CompSci 725, University of Auckland, Oct 2004. b.zdrnja@auckland.ac.nz Abstract Honeypots are security resources which trap malicious activities, so they

More information

Layered Approach of Intrusion Detection System with Efficient Alert Aggregation for Heterogeneous Networks

Layered Approach of Intrusion Detection System with Efficient Alert Aggregation for Heterogeneous Networks Layered Approach of Intrusion Detection System with Efficient Alert Aggregation for Heterogeneous Networks Lohith Raj S N, Shanthi M B, Jitendranath Mungara Abstract Protecting data from the intruders

More information

ADDING NETWORK INTELLIGENCE TO VULNERABILITY MANAGEMENT

ADDING NETWORK INTELLIGENCE TO VULNERABILITY MANAGEMENT ADDING NETWORK INTELLIGENCE INTRODUCTION Vulnerability management is crucial to network security. Not only are known vulnerabilities propagating dramatically, but so is their severity and complexity. Organizations

More information

Intrusion Detection & SNORT. Fakrul Alam fakrul@bdhbu.com

Intrusion Detection & SNORT. Fakrul Alam fakrul@bdhbu.com Intrusion Detection & SNORT Fakrul Alam fakrul@bdhbu.com Sometimes, Defenses Fail Our defenses aren t perfect Patches weren t applied promptly enough Antivirus signatures not up to date 0- days get through

More information

Applying Internal Traffic Models to Improve Identification of High Fidelity Cyber Security Events

Applying Internal Traffic Models to Improve Identification of High Fidelity Cyber Security Events Applying Internal Traffic Models to Improve Identification of High Fidelity Cyber Security Events Abstract Effective Security Operations throughout both DoD and industry are requiring and consuming unprecedented

More information

How to build and use a Honeypot. Ralph Edward Sutton, Jr. DTEC 6873 Section 01

How to build and use a Honeypot. Ralph Edward Sutton, Jr. DTEC 6873 Section 01 How to build and use a Honeypot By Ralph Edward Sutton, Jr DTEC 6873 Section 01 Abstract Everybody has gotten hacked one way or another when dealing with computers. When I ran across the idea of a honeypot

More information

Network Instruments white paper

Network Instruments white paper Network Instruments white paper USING A NETWORK ANALYZER AS A SECURITY TOOL Network Analyzers are designed to watch the network, identify issues and alert administrators of problem scenarios. These features

More information

Cisco IPS Tuning Overview

Cisco IPS Tuning Overview Cisco IPS Tuning Overview Overview Increasingly sophisticated attacks on business networks can impede business productivity, obstruct access to applications and resources, and significantly disrupt communications.

More information

Top 5 Essential Log Reports

Top 5 Essential Log Reports Top 5 Essential Log Reports Version 1.0 Contributors: Chris Brenton - Independent Security Consultant - chris@chrisbrenton.org Tina Bird, Security Architect, PGP Corporation Marcus J Ranum, CSO, Tenable

More information

Fuzzy Network Profiling for Intrusion Detection

Fuzzy Network Profiling for Intrusion Detection Fuzzy Network Profiling for Intrusion Detection John E. Dickerson (jedicker@iastate.edu) and Julie A. Dickerson (julied@iastate.edu) Electrical and Computer Engineering Department Iowa State University

More information

Virtual Terrain: A Security-Based Representation of a Computer Network

Virtual Terrain: A Security-Based Representation of a Computer Network Virtual Terrain: A Security-Based Representation of a Computer Network Jared Holsopple* a, Shanchieh Yang b, Brian Argauer b a CUBRC, 4455 Genesee St, Buffalo, NY, USA 14225; b Dept. of Computer Engineering,

More information

Effective Intrusion Detection

Effective Intrusion Detection Effective Intrusion Detection A white paper by With careful configuration and management, intrusion detection systems can make a valuable contribution to IT infrastructure security s Global network of

More information

A Firewall Data Log Analysis of Unauthorized and Suspicious Traffic

A Firewall Data Log Analysis of Unauthorized and Suspicious Traffic A Firewall Data Log Analysis of Unauthorized and Suspicious Traffic John Week University of Nevada, Reno United States Email:jweek@weekspace.net Phone: (775) 741-1555 Polina Ivanova University of Nevada,

More information

IBM. Vulnerability scanning and best practices

IBM. Vulnerability scanning and best practices IBM Vulnerability scanning and best practices ii Vulnerability scanning and best practices Contents Vulnerability scanning strategy and best practices.............. 1 Scan types............... 2 Scan duration

More information

co Characterizing and Tracing Packet Floods Using Cisco R

co Characterizing and Tracing Packet Floods Using Cisco R co Characterizing and Tracing Packet Floods Using Cisco R Table of Contents Characterizing and Tracing Packet Floods Using Cisco Routers...1 Introduction...1 Before You Begin...1 Conventions...1 Prerequisites...1

More information

SANS Top 20 Critical Controls for Effective Cyber Defense

SANS Top 20 Critical Controls for Effective Cyber Defense WHITEPAPER SANS Top 20 Critical Controls for Cyber Defense SANS Top 20 Critical Controls for Effective Cyber Defense JANUARY 2014 SANS Top 20 Critical Controls for Effective Cyber Defense Summary In a

More information

An Integrated CyberSecurity Approach for HEP Grids. Workshop Report. http://hpcrd.lbl.gov/hepcybersecurity/

An Integrated CyberSecurity Approach for HEP Grids. Workshop Report. http://hpcrd.lbl.gov/hepcybersecurity/ An Integrated CyberSecurity Approach for HEP Grids Workshop Report http://hpcrd.lbl.gov/hepcybersecurity/ 1. Introduction The CMS and ATLAS experiments at the Large Hadron Collider (LHC) being built at

More information

AlienVault. Unified Security Management (USM) 5.x Policy Management Fundamentals

AlienVault. Unified Security Management (USM) 5.x Policy Management Fundamentals AlienVault Unified Security Management (USM) 5.x Policy Management Fundamentals USM 5.x Policy Management Fundamentals Copyright 2015 AlienVault, Inc. All rights reserved. The AlienVault Logo, AlienVault,

More information

From Network Security To Content Filtering

From Network Security To Content Filtering Computer Fraud & Security, May 2007 page 1/10 From Network Security To Content Filtering Network security has evolved dramatically in the last few years not only for what concerns the tools at our disposals

More information

INTRUSION DETECTION SYSTEM (IDS) by Kilausuria Abdullah (GCIH) Cyberspace Security Lab, MIMOS Berhad

INTRUSION DETECTION SYSTEM (IDS) by Kilausuria Abdullah (GCIH) Cyberspace Security Lab, MIMOS Berhad INTRUSION DETECTION SYSTEM (IDS) by Kilausuria Abdullah (GCIH) Cyberspace Security Lab, MIMOS Berhad OUTLINE Security incident Attack scenario Intrusion detection system Issues and challenges Conclusion

More information

Improving the Database Logging Performance of the Snort Network Intrusion Detection Sensor

Improving the Database Logging Performance of the Snort Network Intrusion Detection Sensor -0- Improving the Database Logging Performance of the Snort Network Intrusion Detection Sensor Lambert Schaelicke, Matthew R. Geiger, Curt J. Freeland Department of Computer Science and Engineering University

More information

Flow-based detection of RDP brute-force attacks

Flow-based detection of RDP brute-force attacks Flow-based detection of RDP brute-force attacks Martin Vizváry vizvary@ics.muni.cz Institute of Computer Science Masaryk University Brno, Czech Republic Jan Vykopal vykopal@ics.muni.cz Institute of Computer

More information

Distributed Denial of Service Attack Tools

Distributed Denial of Service Attack Tools Distributed Denial of Service Attack Tools Introduction: Distributed Denial of Service Attack Tools Internet Security Systems (ISS) has identified a number of distributed denial of service tools readily

More information

Firewalls, Tunnels, and Network Intrusion Detection. Firewalls

Firewalls, Tunnels, and Network Intrusion Detection. Firewalls Firewalls, Tunnels, and Network Intrusion Detection 1 Firewalls A firewall is an integrated collection of security measures designed to prevent unauthorized electronic access to a networked computer system.

More information

The SIEM Evaluator s Guide

The SIEM Evaluator s Guide Using SIEM for Compliance, Threat Management, & Incident Response Security information and event management (SIEM) tools are designed to collect, store, analyze, and report on log data for threat detection,

More information

An Anomaly-Based Method for DDoS Attacks Detection using RBF Neural Networks

An Anomaly-Based Method for DDoS Attacks Detection using RBF Neural Networks 2011 International Conference on Network and Electronics Engineering IPCSIT vol.11 (2011) (2011) IACSIT Press, Singapore An Anomaly-Based Method for DDoS Attacks Detection using RBF Neural Networks Reyhaneh

More information

How to Painlessly Audit Your Firewalls

How to Painlessly Audit Your Firewalls W h i t e P a p e r How to Painlessly Audit Your Firewalls An introduction to automated firewall compliance audits, change assurance and ruleset optimization May 2010 Executive Summary Firewalls have become

More information

Name. Description. Rationale

Name. Description. Rationale Complliiance Componentt Description DEEFFI INITION Network-Based Intrusion Detection Systems (NIDS) Network-Based Intrusion Detection Systems (NIDS) detect attacks by capturing and analyzing network traffic.

More information

Host Hardening. Presented by. Douglas Couch & Nathan Heck Security Analysts for ITaP 1

Host Hardening. Presented by. Douglas Couch & Nathan Heck Security Analysts for ITaP 1 Host Hardening Presented by Douglas Couch & Nathan Heck Security Analysts for ITaP 1 Background National Institute of Standards and Technology Draft Guide to General Server Security SP800-123 Server A

More information

Firewalls. Chapter 3

Firewalls. Chapter 3 Firewalls Chapter 3 1 Border Firewall Passed Packet (Ingress) Passed Packet (Egress) Attack Packet Hardened Client PC Internet (Not Trusted) Hardened Server Dropped Packet (Ingress) Log File Internet Border

More information

How To Connect Log Files To A Log File On A Network With A Network Device (Network) On A Computer Or Network (Network Or Network) On Your Network (For A Network)

How To Connect Log Files To A Log File On A Network With A Network Device (Network) On A Computer Or Network (Network Or Network) On Your Network (For A Network) SIEM FOR BEGINNERS EVERYTHING YOU WANTED TO KNOW ABOUT LOG MANAGEMENT BUT WERE AFRAID TO ASK www.alienvault.com A Rose By Any Other Name SLM/LMS, SIM, SEM, SEC, SIEM Although the industry has settled on

More information

Intrusion Detection. Tianen Liu. May 22, 2003. paper will look at different kinds of intrusion detection systems, different ways of

Intrusion Detection. Tianen Liu. May 22, 2003. paper will look at different kinds of intrusion detection systems, different ways of Intrusion Detection Tianen Liu May 22, 2003 I. Abstract Computers are vulnerable to many threats. Hackers and unauthorized users can compromise systems. Viruses, worms, and other kinds of harmful code

More information

Machine-to-Machine Exchange of Cyber Threat Information: a Key to Mature Cyber Defense

Machine-to-Machine Exchange of Cyber Threat Information: a Key to Mature Cyber Defense Machine-to-Machine Exchange of Cyber Threat Information: a Key to Mature Cyber Defense By: Daniel Harkness, Chris Strasburg, and Scott Pinkerton The Challenge The Internet is an integral part of daily

More information

AlienVault Unified Security Management (USM) 4.x-5.x. Deployment Planning Guide

AlienVault Unified Security Management (USM) 4.x-5.x. Deployment Planning Guide AlienVault Unified Security Management (USM) 4.x-5.x Deployment Planning Guide USM 4.x-5.x Deployment Planning Guide, rev. 1 Copyright AlienVault, Inc. All rights reserved. The AlienVault Logo, AlienVault,

More information

Alarm Clustering for Intrusion Detection Systems in Computer Networks

Alarm Clustering for Intrusion Detection Systems in Computer Networks Alarm Clustering for Intrusion Detection Systems in Computer Networks Giorgio Giacinto, Roberto Perdisci, Fabio Roli Department of Electrical and Electronic Engineering, University of Cagliari Piazza D

More information

The Reverse Firewall: Defeating DDOS Attacks Emanating from a Local Area Network

The Reverse Firewall: Defeating DDOS Attacks Emanating from a Local Area Network Pioneering Technologies for a Better Internet Cs3, Inc. 5777 W. Century Blvd. Suite 1185 Los Angeles, CA 90045-5600 Phone: 310-337-3013 Fax: 310-337-3012 Email: info@cs3-inc.com The Reverse Firewall: Defeating

More information

Firewalls, Tunnels, and Network Intrusion Detection

Firewalls, Tunnels, and Network Intrusion Detection Firewalls, Tunnels, and Network Intrusion Detection 1 Part 1: Firewall as a Technique to create a virtual security wall separating your organization from the wild west of the public internet 2 1 Firewalls

More information

What a Vulnerability Assessment Scanner Can t Tell You. Leveraging Network Context to Prioritize Remediation Efforts and Identify Options

What a Vulnerability Assessment Scanner Can t Tell You. Leveraging Network Context to Prioritize Remediation Efforts and Identify Options White paper What a Vulnerability Assessment Scanner Can t Tell You Leveraging Network Context to Prioritize Remediation Efforts and Identify Options november 2011 WHITE PAPER RedSeal Networks, Inc. 3965

More information

Radware s Smart IDS Management. FireProof and Intrusion Detection Systems. Deployment and ROI. North America. International. www.radware.

Radware s Smart IDS Management. FireProof and Intrusion Detection Systems. Deployment and ROI. North America. International. www.radware. Radware s Smart IDS Management FireProof and Intrusion Detection Systems Deployment and ROI North America Radware Inc. 575 Corporate Dr. Suite 205 Mahwah, NJ 07430 Tel 888 234 5763 International Radware

More information

Chapter 9 Firewalls and Intrusion Prevention Systems

Chapter 9 Firewalls and Intrusion Prevention Systems Chapter 9 Firewalls and Intrusion Prevention Systems connectivity is essential However it creates a threat Effective means of protecting LANs Inserted between the premises network and the to establish

More information

CSCI 4250/6250 Fall 2015 Computer and Networks Security

CSCI 4250/6250 Fall 2015 Computer and Networks Security CSCI 4250/6250 Fall 2015 Computer and Networks Security Network Security Goodrich, Chapter 5-6 Tunnels } The contents of TCP packets are not normally encrypted, so if someone is eavesdropping on a TCP

More information

White Paper. Intrusion Detection Deploying the Shomiti Century Tap

White Paper. Intrusion Detection Deploying the Shomiti Century Tap White Paper Intrusion Detection Deploying the Shomiti Century Tap . Shomiti Tap Deployment Purpose of this Paper The scalability of Intrusion Detection Systems (IDS) is often an issue when deploying an

More information

KEITH LEHNERT AND ERIC FRIEDRICH

KEITH LEHNERT AND ERIC FRIEDRICH MACHINE LEARNING CLASSIFICATION OF MALICIOUS NETWORK TRAFFIC KEITH LEHNERT AND ERIC FRIEDRICH 1. Introduction 1.1. Intrusion Detection Systems. In our society, information systems are everywhere. They

More information

CHAPETR 3. DISTRIBUTED DEPLOYMENT OF DDoS DEFENSE SYSTEM

CHAPETR 3. DISTRIBUTED DEPLOYMENT OF DDoS DEFENSE SYSTEM 59 CHAPETR 3 DISTRIBUTED DEPLOYMENT OF DDoS DEFENSE SYSTEM 3.1. INTRODUCTION The last decade has seen many prominent DDoS attack on high profile webservers. In order to provide an effective defense against

More information

Firewalls, IDS and IPS

Firewalls, IDS and IPS Session 9 Firewalls, IDS and IPS Prepared By: Dr. Mohamed Abd-Eldayem Ref.: Corporate Computer and Network Security By: Raymond Panko Basic Firewall Operation 2. Internet Border Firewall 1. Internet (Not

More information

ECE 578 Term Paper Network Security through IP packet Filtering

ECE 578 Term Paper Network Security through IP packet Filtering ECE 578 Term Paper Network Security through IP packet Filtering Cheedu Venugopal Reddy Dept of Electrical Eng and Comp science Oregon State University Bin Cao Dept of electrical Eng and Comp science Oregon

More information

Network Security Monitoring: Looking Beyond the Network

Network Security Monitoring: Looking Beyond the Network 1 Network Security Monitoring: Looking Beyond the Network Ian R. J. Burke: GCIH, GCFA, EC/SA, CEH, LPT iburke@headwallsecurity.com iburke@middlebury.edu February 8, 2011 2 Abstract Network security monitoring

More information

Five Tips to Ensure Data Loss Prevention Success

Five Tips to Ensure Data Loss Prevention Success Five Tips to Ensure Data Loss Prevention Success A DLP Experts White Paper January, 2013 Author s Note The content of this white paper was developed independently of any vendor sponsors and is the sole

More information

Flexible Web Visualization for Alert-Based Network Security Analytics

Flexible Web Visualization for Alert-Based Network Security Analytics Flexible Web Visualization for Alert-Based Network Security Analytics Lihua Hao 1, Christopher G. Healey 1, Steve E. Hutchinson 2 1 North Carolina State University, 2 U.S. Army Research Laboratory lhao2@ncsu.edu

More information

Guideline on Auditing and Log Management

Guideline on Auditing and Log Management CMSGu2012-05 Mauritian Computer Emergency Response Team CERT-MU SECURITY GUIDELINE 2011-02 Enhancing Cyber Security in Mauritius Guideline on Auditing and Log Management National Computer Board Mauritius

More information

Customer Service Description Next Generation Network Firewall

Customer Service Description Next Generation Network Firewall Customer Service Description Next Generation Network Firewall Interoute, Walbrook Building, 195 Marsh Wall, London, E14 9SG, UK Tel: +800 4683 7681 Email: info@interoute.com Interoute Communications Limited

More information

CSE331: Introduction to Networks and Security. Lecture 17 Fall 2006

CSE331: Introduction to Networks and Security. Lecture 17 Fall 2006 CSE331: Introduction to Networks and Security Lecture 17 Fall 2006 Announcements Project 2 is due next Weds. Homework 2 has been assigned: It's due on Monday, November 6th. CSE331 Fall 2004 2 Summary:

More information

2 Technologies for Security of the 2 Internet

2 Technologies for Security of the 2 Internet 2 Technologies for Security of the 2 Internet 2-1 A Study on Process Model for Internet Risk Analysis NAKAO Koji, MARUYAMA Yuko, OHKOUCHI Kazuya, MATSUMOTO Fumiko, and MORIYAMA Eimatsu Security Incidents

More information

83-10-40 Firewalls: An Effective Solution for Internet Security E. Eugene Schultz Payoff

83-10-40 Firewalls: An Effective Solution for Internet Security E. Eugene Schultz Payoff 83-10-40 Firewalls: An Effective Solution for Internet Security E. Eugene Schultz Payoff Firewalls are an effective method of reducing the possibility of network intrusion by attackers. The key to successful

More information

Rule 4-004M Payment Card Industry (PCI) Monitoring, Logging and Audit (proposed)

Rule 4-004M Payment Card Industry (PCI) Monitoring, Logging and Audit (proposed) Version: Modified By: Date: Approved By: Date: 1.0 Michael Hawkins October 29, 2013 Dan Bowden November 2013 Rule 4-004M Payment Card Industry (PCI) Monitoring, Logging and Audit (proposed) 01.1 Purpose

More information

Quick Start for Network Agent. 5-Step Quick Start. What is Network Agent?

Quick Start for Network Agent. 5-Step Quick Start. What is Network Agent? What is Network Agent? The Websense Network Agent software component uses sniffer technology to monitor all of the internet traffic on the network machines that you assign to it. Network Agent filters

More information

IP Link Best Practices for Network Integration and Security. Introduction...2. Passwords...4 ACL...5 VLAN...6. Protocols...6. Conclusion...

IP Link Best Practices for Network Integration and Security. Introduction...2. Passwords...4 ACL...5 VLAN...6. Protocols...6. Conclusion... IP Link Best Practices for Network Integration and Security Table of Contents Introduction...2 Passwords...4 ACL...5 VLAN...6 Protocols...6 Conclusion...9 Abstract Extron IP Link technology enables A/V

More information

Host-based Intrusion Prevention System (HIPS)

Host-based Intrusion Prevention System (HIPS) Host-based Intrusion Prevention System (HIPS) White Paper Document Version ( esnhips 14.0.0.1) Creation Date: 6 th Feb, 2013 Host-based Intrusion Prevention System (HIPS) Few years back, it was relatively

More information

New Era in Cyber Security. Technology Development

New Era in Cyber Security. Technology Development New Era in Cyber New Era in Cyber Security Security Technology Technology Development Development Combining the Power of the Oil and Gas Industry, DHS, and the Vendor Community to Combat Cyber Security

More information

Module II. Internet Security. Chapter 7. Intrusion Detection. Web Security: Theory & Applications. School of Software, Sun Yat-sen University

Module II. Internet Security. Chapter 7. Intrusion Detection. Web Security: Theory & Applications. School of Software, Sun Yat-sen University Module II. Internet Security Chapter 7 Intrusion Detection Web Security: Theory & Applications School of Software, Sun Yat-sen University Outline 7.1 Threats to Computer System 7.2 Process of Intrusions

More information

Cognitive and Organizational Challenges of Big Data in Cyber Defense

Cognitive and Organizational Challenges of Big Data in Cyber Defense Cognitive and Organizational Challenges of Big Data in Cyber Defense Nathan Bos & John Gersh Johns Hopkins University Applied Laboratory nathan.bos@jhuapl.edu, john.gersh@jhuapl.edu The cognitive and organizational

More information

AlienVault. Unified Security Management (USM) 5.1 Running the Getting Started Wizard

AlienVault. Unified Security Management (USM) 5.1 Running the Getting Started Wizard AlienVault Unified Security Management (USM) 5.1 Running the Getting Started Wizard USM v5.1 Running the Getting Started Wizard, rev. 2 Copyright 2015 AlienVault, Inc. All rights reserved. The AlienVault

More information

White Paper. What Auditors Want Database Auditing. 5 Key Questions Auditors Ask During a Database Compliance Audit

White Paper. What Auditors Want Database Auditing. 5 Key Questions Auditors Ask During a Database Compliance Audit 5 Key Questions Auditors Ask During a Database Compliance Audit White Paper Regulatory legislation is increasingly driving the expansion of formal enterprise audit processes to include information technology

More information

THE ROLE OF IDS & ADS IN NETWORK SECURITY

THE ROLE OF IDS & ADS IN NETWORK SECURITY THE ROLE OF IDS & ADS IN NETWORK SECURITY The Role of IDS & ADS in Network Security When it comes to security, most networks today are like an egg: hard on the outside, gooey in the middle. Once a hacker

More information

Radware s Behavioral Server Cracking Protection

Radware s Behavioral Server Cracking Protection Radware s Behavioral Server Cracking Protection A DefensePro Whitepaper By Renaud Bidou Senior Security Specialist,Radware October 2007 www.radware.com Page - 2 - Table of Contents Abstract...3 Information

More information

The Importance of Cybersecurity Monitoring for Utilities

The Importance of Cybersecurity Monitoring for Utilities The Importance of Cybersecurity Monitoring for Utilities www.n-dimension.com Cybersecurity threats against energy companies, including utilities, have been increasing at an alarming rate. A comprehensive

More information

The Truth about False Positives

The Truth about False Positives An ISS Technical White Paper The Truth about False Positives 6303 Barfield Road Atlanta, GA 30328 Tel: 404.236.2600 Fax: 404.236.2626 Overview In the security industry, many security analysts remark that

More information

Chapter 15. Firewalls, IDS and IPS

Chapter 15. Firewalls, IDS and IPS Chapter 15 Firewalls, IDS and IPS Basic Firewall Operation The firewall is a border firewall. It sits at the boundary between the corporate site and the external Internet. A firewall examines each packet

More information

Vulnerability Management

Vulnerability Management Vulnerability Management Buyer s Guide Buyer s Guide 01 Introduction 02 Key Components 03 Other Considerations About Rapid7 01 INTRODUCTION Exploiting weaknesses in browsers, operating systems and other

More information

SysPatrol - Server Security Monitor

SysPatrol - Server Security Monitor SysPatrol Server Security Monitor User Manual Version 2.2 Sep 2013 www.flexense.com www.syspatrol.com 1 Product Overview SysPatrol is a server security monitoring solution allowing one to monitor one or

More information

Firewalls. Test your Firewall knowledge. Test your Firewall knowledge (cont) (March 4, 2015)

Firewalls. Test your Firewall knowledge. Test your Firewall knowledge (cont) (March 4, 2015) s (March 4, 2015) Abdou Illia Spring 2015 Test your knowledge Which of the following is true about firewalls? a) A firewall is a hardware device b) A firewall is a software program c) s could be hardware

More information

Outline Intrusion Detection CS 239 Security for Networks and System Software June 3, 2002

Outline Intrusion Detection CS 239 Security for Networks and System Software June 3, 2002 Outline Intrusion Detection CS 239 Security for Networks and System Software June 3, 2002 Introduction Characteristics of intrusion detection systems Some sample intrusion detection systems Page 1 Page

More information

Network device management solution

Network device management solution iw Management Console Network device management solution iw MANAGEMENT CONSOLE Scalability. Reliability. Real-time communications. Productivity. Network efficiency. You demand it from your ERP systems

More information