Voice over Internet Protocol (VoIP) technology has come of age and is quickly gaining momentum on Broadband networks. VoIP packetizes phone calls through the same routes used by network and Internet traffic and is consequently prone to the same cyber threats that plague data networks today. It presents lower cost and greater flexibility for a venture but presents considerable security challenges. Many solutions for VoIP security are projected, however these solutions should take into account the real-time constriction of voice service and their methods be supposed to address probable attacks and overhead related with it.

One of these solutions is to make use of Firewalls, which implement a security strategy by examining and straining traffic arriving or leaving from a protected network. This is normally done by evaluating an incoming packet to a set of policies and performing the corresponding rule action, which is accept or reject. Undesirably packet examinations can require considerable interruptions on traffic due to the difficulty and size of policies. Consequently, improving firewall performance is significant for the VoIP networks. In this paper, we propose a new firewall deign that is able to dynamically update firewall policy based on Neural Network and achieve packet examinations under rising traffic loads, higher traffic speeds, and stringent QoS necessities.

The design consists of several firewalls configured in parallel that jointly impose a defense strategy. Every firewall outfits part of the rule and incoming packets is processed through all the firewalls concurrently. Once the neural network is trained, it continuously updates the firewall policy using the selected parameters to perform its evaluation. Since many firewalls are utilized to process each packet, the proposed parallel firewall system has considerably lower delays and a higher throughput than other firewalls.


Voice over IP the transmission of voice over traditional packet-switched IP networks is one of the hottest trends in telecommunications. Although most computers can provide VoIP and many offer VoIP applications, the term “voice over IP” is typically associated with equipment that lets users dial telephone numbers and communicate with parties on the other end who have a VoIP system or a traditional analog telephone. (The sidebar, “Current voice-over-IP products,” describes some of the products on the market today.) As with any new technology, VoIP introduces both opportunities and problems. It offers lower cost and greater flexibility for an enterprise but presents significant security challenges.

As with any new technology, VoIP introduces both opportunities and problems.. Security administrators might assume that because digitized voice travels in packets, they can simply plug VoIP components into their already secured networks and get a stable and secure voice network. Quality of service (QoS) is fundamental to a VoIP networks operation. A VoIP application is much more sensitive to delays than its traditional data counterparts. Latency turns traditional security measures into double-edged swords for VoIP.

Tools such as encryption and firewall protection can help secure the network, but they also produce significant delay. Latency isn't just a QoS issue, but also a security issue because it increases the system's susceptibility to denial-of-service attacks. To succeed in a VoIP network, a DoS attack need not completely shut down the system, but only delay voice packets for a fraction of a second. The necessary impediment is even less when latency-producing security devices are slowing down traffic.

As described in the introduction, parallelization offers a Scalable technique for improving the performance of network firewalls. Using this approach an array of m firewalls processes packets in parallel. However, the two designs depicted in differ based on what is distributed: packets or rules. The design was Consisted of multiple identical firewalls connected in parallel, each firewall j in the system implements a local policy Rj where Rj = R. Arriving packets are distributed across the firewalls for processing (one packet is sent to one firewall), allowing different packets to be processed in parallel. Since each packet is processed using the policy Rj = R, policy integrity is maintained.

A neural network is a group of interconnected nodes. The well-known example is the human brain, the most complicated and difficult neural network. We can make very fast and reliable choice in portion of a second. In the face of the clear neatness of usual thinking, outcome are usually not-white and -black or binary, but quite engage a broad diversity of alert and secreted inputs, we have an wonderful facility to recognize well-known patterns as well as extraordinary patterns more or less directly, the neural network approach effort to reproduce the way humans visually the usual consumer speedily studies to identify spam from correct connection.

The reason for this is generally since we illustration our brains both on reason to a broad variety of message content and the brain learns to create lightning-fast, very exact guess. The capacity of utilizing packet changed networks as a transmit standard for real-time tone of voice connections has drawn broad awareness among both research and possible communities alike.

The current progress in speech conventions and high speed information communication technology hold up the notice in equipment such as voice over Internet protocol (VoIP), the mathematical character of information interchange and the energetic routing method engaged in packet-switched networks outcomes in an unbalanced network delay (jitter) practiced by IP packets. [chris miller].

Although a data-parallel firewall can achieve higher throughput than a traditional (single machine) firewall, it suffers from two major disadvantages. First, stateful inspection requires all traffic from a certain connection or exchange to traverse the same firewall. Successful connection tracking is difficult to perform at high speeds using the data-parallel approach Second, distributing packets is only beneficial when each firewall in the array has a significant amount of traffic to process (never idle), which only occurs under high traffic loads.

In order to understand parity in a carrier network maintaining secrecy, the studying techniques to know excluded traffic from partial information, such as the header information and show pattern of a series of packets. The propose a traffic credit technique for a direct request which uses mathematical information such as incidence of packet coming.

This method is to be used for stop idea by recognize traffic generate by not only VoIP but video request as well. By using this method, travel that is clearly mediator excluded is not needed, the quality of traffic that is classified into best services, such as urgent situation message and moving sharing, is certain, and, for best effort services, suitable operation are perform so that capital can not be busy by a few edge, so as to understand fairness in symbol services. This advance is to applications that generate traffic from the presentation of the traffic. It can be underground into the following three types regarding the granularity of the observed traffic. [Toshiya Okabe Tsutomu Kitamura 2006].

Transaction-level behavior

This approach is a system focus on the skin of an application-level action, such as an HTTP request message, and its response, an HTTP response message or MAIL message. With this progress, a request is indirect from the change patterns, the size of each message. A method to order maintain by the time-series changes in the size of messages. These techniques are useful for sense a signal protocol, but are not suitable for discovery of real-time message traffic whose features are boring and last for a fairly long time. [Toshiya Okabe Tsutomu Kitamura 2006]

Flow-level behavior

This method is to make out an application from mathematical information such as the inter-arrival time, period of the run, packet size. Here, a run is defined as a sequence of packets having a common source address, source port, destination address, destination port and transport protocol. It is extract skin of size data message request flows, such as HTTP, FTP and SMTP graceful from side to side a network, in order to create workload for a network simulator and classify traffic into three lessons, bulk data message such as FTP, informal message. [Toshiya Okabe Tsutomu Kitamura 2006]

Packet-level behavior

This is techniques that identify a request from the header or load of a single packet. A group method mainly based on port facts has been used but its efficiency has been lost due to the arrival of P2P applications that illegitimately use chance port numbers and port numbers for HTTP to traverse a Firewall. [Toshiya Okabe Tsutomu Kitamura 2006].

Average packet length and variation

The result of extract skin correlated to the normal packet size and difference in packet size. Difference in packet size here indicates the number of types of packet size for a request whose packet size is fixed. It is the result of take out the skin of voice applications. The packet size of the voice application is lesser than that of the other application. [ Takayuki Shizuno 2006]

As with any new knowledge, VoIP introduce both opportunity and problems. It offers lesser cost and greater give for an project but presents significant security challenge. Security administrator strength suppose that because digitized voice actions in packets, the plug VoIP mechanism into their previously protected networks and get a stable and secure voice network address translation (NAT), and most VoIP mechanism have counterpart in data network, VoIP's presentation stress mean you must extra ordinary network software and hardware with special VoIP mechanism.

Packet network depend on many configurable bound: IP and physical addresses of say terminal of routers and firewalls. VoIP networks add specific software, to place and route calls. Many network bound are recognized with passion each time a network part is restart or when a VoIP phone is restart or added to the network.. So many nodes in a VoIP network have dynamically configurable bound; But VoIP systems have much stricter presentation constraint than data networks with important implication for security. [ Takayuki Shizuno 2006].

Quality-of-service issues

Quality of service is basic to a VoIP network's process. A VoIP request is much more responsive to delay than its customary data matching part. In the VoIP language, this is the latency problem. Latency turns conventional safety measured. Tools such as encryption and firewall defense can help secure the system, but they also set up important delay. Latency isn't just a QoS issue, but also a safety issue because it increases the system weakness to denial-of-service attacks.

To do well in a VoIP network, a DoS attack need not totally shut down the system, but only delay voice packets for a part of a second. The necessary let is even less when latency-producing security devices are slowing down traffic. Another QoS issue jitter, refers to no uniform delays that can cause packets to turn up and be process out of series. The Real-Time Transport Protocol (RTP), which is used to move voice media, so packets received out of order can't be reassembled at the move level, but must be rearrange at the request level, introduce major above your head. When packets turn up in order, high jitter causes them to arrive at their target in spurts.

To control jitter, network expensive can use buffers and implement QoS-supporting network elements that let VoIP packets when larger data packets are listed in front of them. The buffer can use one of several plans to resolve when to let go voice data, counting several scheme that adapt the payout time also encompass packet loss. In addition to the usual packet loss issue related with data networks, even VoIP packets that reach their target can be make useless by latency and jitter. [thomas j. walsh and d. richard kuhn ].

Project Background

Neural network is the bury order growing fast in current years. It is jointly of a massive deal of easy giving out units of neuron with providing connect as a neural network. It can replicate the information distribution task of human being brain, with huge talent of nonlinear estimate, consecutively storage, large-scale similar development, and self-training lessons. The information distribution in the neural network is recognizing by the communication between the neurons, and the storage of data and in progression as increase physical interconnection of the network parts. [, a. shelestov, v. pasechnik, a. sidorenko, n. kussul , 2006].

A parallel firewall (also called a load-balancing firewall) is a scalable approach for increasing the speed of inspecting network traffic. As seen in figure .the system consists of multiple identical firewalls connected in parallel. Each firewall in the system implements the complete security policy and arriving packets are distributed across the firewalls such that only one firewall processes any given packet. How the load-balancing algorithm distributes packets is vital to the system and typically implemented as a high-speed switch in commercial products.

Although parallel firewalls achieve a higher throughput than traditional firewalls and have a redundant design, the performance benefit is only evident under high traffic loads. Furthermore, stateful inspection requires all traffic from a certain connection or exchange to traverse the same firewall, which is difficult to perform at high speeds. This paper introduces a new scalable parallel firewall architecture designed for increasing network speeds and traffic loads. The design consists of multiple firewalls where each firewall implements only a portion of the security policy.

Since the policy is divided across the firewalls, rule distribution guidelines are provided that maintains integrity, ensuring the new parallel design and a traditional single firewall always reach the same decision. Unlike the previous parallel design, When a packet arrives to the new architecture it is processed by every firewall in parallel, thus the processing time required per packet is reduced.

Simulation results for the new architecture (consisting of four firewalls) yielded a 74% reduction in processing time as compared to other parallel firewall designs. Furthermore, the proposed architecture can provide stateful inspections since a packet is processed by every firewall. Therefore, the new parallel design is a scalable solution that can offer better performance and more capabilities than other designs.

In list-based rule symbol, when packets arrive at a firewall, it is in sequence check against the system in the rule list until a match is found or attainment the end of the list. Then, the parallel action is applied to mass or pass the packet. To make the policy complete of match is always found for each packet, the computational difficulty of the sort process depends on the length of rule as the depth of result a matched rule in the rule list. Apply more composite policy can result in major traffic wait which is not only a presentation block in high speed environment but also can make it weaker to rejection of service attacks.

Moreover, attractive the filter time is more difficult for multimedia applications that require firm quality of service promise. Although hardware solutions can very much decrease the packet giving out time, they are costly for large policy and improvement hardware may not be suitable in inheritance systems. on the other hand, better data structure for inner policy symbol and better search mechanism have been planned to provide relatively and effective solution to benefit on hand hardware systems.

The rules are group to allow multidimensional search by at once eliminate multiple rules with few comparison. While tries have shown great agree in improving the search time, the storage condition and difficulty in maintain try and policy honesty increases as more rules . Moreover, a policy trie does not take into report the traffic personality. In a method for trie sorting is proposed that sustain the policy honesty while reorder rules for unreliable traffic situation. The number of contrast as compare to the original trie.

Traffic-aware optimization of list-based firewalls has been addressed in rules are assigned matching probability that depend on the traffic information. Other firewall models have been planned to signify and analyze policy whether for central or distributed firewall architectures with main focus on identify rule conflicts and variance; Wireless local (WLAN) and wireless personal (WPAN) area networks are being used increasingly to implement VoIP forces. The main drives for using these architectures are user mobility. Behind realible real-time repair is one of the major concerns for generally use of VoIP in these wireless IPbased networks and safety is now getting the notice of researchers. The security and efficiency are consisting requirements. [El-Sayed M. El-Alfy and Shokri Z. Selim 2007]

Literature Review

  • Introduction:

Literature Review is the process of finding information for help on searching for resources on the Internet. Reading intensively in the chosen topic area is essential, but the task can prove daunting if they do not approach it in a systematic way. The continuous number of high-profile Internet security breeches reported in the mass media shows that despite an emphasis on security processes that there is still a gap between theory and practice.

Not only is there a need to develop better software engineering processes but also theoretical security improvements need to find their way into real systems. Software design patterns are defined as “descriptions of communicating objects and classes that are customized to solve a general design problem in a particular context”. As software design patterns have proven their value in the development of production software, they are a promising new approach to help in both the theoretical development and practical implementation of better security processes.

First, many/most software developers have only a limited knowledge of security processes and patterns are a proven way to improve their understanding. Second, patterns work against “reinventing-the-wheel” to promote learning best practices from the larger community to save time, effort, and money with easily accessible and validated examples. Third, code can be reused since the same security patterns arise in many different contexts

  • Investigating existing resources in our area of research will generally cover three areas:
  • "Exploratory investigations, as part of the development and evaluation of possible topics in an area
  • Investigation is some depth, sufficient to support a formal research and dissertation proposal
  • Complete research that is described in the 'literature / research' section of the dissertation." [from Writing the Doctoral Dissertation, To Author names]

2. Related Research Work Available:

Wireless local (WLAN) and wireless personal (WPAN) area networks are being used progressively to implement VoIP services. The main motivation for using these architectures are user mobility, setup flexibility, increasing transmission rate and low costs, despite this convergence depends on the answers of several technical problems

Supporting reliable real-time service is one of the major concerns for widely deployment of VoIP in these ireless IPbased networks and security is now receiving the attention of researchers. The problem of offering security to WLAN and WPAN is that security does not come for free and, security and efficiency are conflicting requirements. The introduction of a security mechanism such as the IPSec encryption-engine to overcome these issues impacts directly in the speech quality of established calls and in the channel capacity.

Moreover, largely deployed radio technology standards as IEEE 802.11 and Bluetooth used to achieve wireless connectivity have several constraints when delivering real-time traffic, as transmission errors at the channel, introducing delay and loss which with security mechanisms impact can lead to low quality VoIP calls. Although these technologies offer some security mechanisms, they have some flaws which need to be addressed by an additional level of security. In this paper we focus on the IPSec protocol to achieve the data secrecy due to its widely deployment and implementation of many encryption algorithms.

During final decades information technology founded on the computer networks take part in an essential role in different areas of human being action. Troubles of huge importance are assigned on them, such as maintenance, communication and mechanization of information processing. The safety level of processed information is able to differ from private and viable to military and state secret.

Herewith the destruction of the information secrecy, reliability and accessibility may cause the spoil to its proprietor and contain important unattractive consequences. Hence the trouble of information safety is concerned. Many associations and companies expand safety facilities that need important aids. In additional, the impracticality of creating wholly protected system is a recognized fact – it will always hold faults and «gaps» in its understanding.

To guard computer systems such familiarized mechanisms as classification and verification, methodologies of the delimitation and limit of the access to data and cryptographic techniques are applied.

But they hold following drawbacks:

• Disclosure from interior users with spiteful purpose;

• Complexity in access separation caused by data sources globalization, which cleans

away difference between "personal" and "foreign" topics of the system;

• Diminution of efficiency and communication complexity by reason of methods for

access control to the sources, for occasion, in e-commerce;

• Effortlessness of passwords description by crating arrangements of simple users' relations.

Hence classification and audit systems are utilized beside with these methods. between them are interruption.

Intrusion Detection Systems (IDS).

IDS are generally separated to systems detecting previously identified attacks (mishandling exposure systems) and variance exposure systems registering the life cycle differences of the computer system from its usual (distinctive) action. Besides, IDS are divided to network-based and host-based category by data source. Network-based IDS examine network dataflow, caring its members, almost not moving the output of their work. Network-based systems do not utilize data about progression from divide workstation.

A firewall is a mixture of hardware and software used to put into practice a security policy leading the flow of network traffic between two or more networks. In its simplest form, a firewall acts as a safety barrier to control traffic and manage links between internal and external network hosts. The actual means by which this is able varies and ranges from packet sort and proxy service to stateful examination methods.

A more difficult firewall may hide the topology of the network it is employed to keep, Firewalls have recognized to be useful in trade with a large number of pressure that create from outer a network. They are becoming ever-present and necessary to the action of the network. The constant growth of the Internet, coupled with the increasing difficulty of attacks, however, is placing further stress and difficulty on firewalls design and management. . [ Subrata Acharya, Jia Wang, Albert Greenberg 2006]

Furthermore, the need to deal with large set of varied safety policy and rules impose additional load on firewalls, thus depiction the presentation of the firewall highly serious to enforce the network safety policy. In this context, the defense that a firewall provides only the policies it is configured to execute, but evenly importantly the speed at which it enforces these policy. Under attack or deep load, firewalls can simply become a bottleneck. As the network size, bandwidth, and giving out power of networked hosts carry on increasing, there is a high demand for optimizing firewall operation for improved performance. [ Subrata Acharya, Jia Wang, Albert Greenberg 2006]

Multi-dimensional firewall research group of people to focus on mounting various optimizations to make firewalls more resourceful and steady. In spite of significant progress in the design of firewalls, the techniques for firewall optimization remains static and fail to get used to to the always varying dynamics of the network. This is frequently due to their failure to take into account the traffic individuality by the firewall, such as source and purpose, service requests and the resultant action taken by the firewall in reply to these requests.

Moreover, current firewall designs do not support adaptive difference discovery and counter measure device. As a result, they run the risk to become unbalanced under attack. The object of this paper is to address the above failing and develop a sound and effective toolset to hasten firewall operation and adapt its performance to the dynamically altering network traffic individuality.

Achieve this goal, however is tough, as the number of policy and safety rules a firewall has to enforce for enterprise network. In addition, there is a need for preserve high policy addition. This is further compounded by the limited resources of firewalls relation to the increased ability of the network to process and forward traffic at very high speed. [ Subrata Acharya, Jia Wang, Albert Greenberg 2006]


Network Firewall Parallelization

Firewall parallelization is a scalable move toward for attaining the speed of system traffic assessment [Carsten Benecke,1999] necessary for improved network paces and traffic loads. In this section data parallel [Carsten Benecke,1999] and function parallel designs [Errin W. Fulp,2002] for parallel firewalls are explained. Similar to their distributed computing descriptions necessitate, data parallel partition the information set across the array of firewalls, while function parallel distributes the work set across the array of firewalls.

The data parallel technique is a scalable substitute to a single firewall that permits for better throughput potentials. Function parallel techniques can diminish the dispensation time necessary on any firewall node yielding enhanced presentation. In addition, unlike data parallel designs, the proposed function parallel architecture can offer stateful assessments. This proposal shows that function parallel designs are scalable solution that can offer better performance and more facilities than other designs.

Parallel Firewall Architecture:

Function Parallel Firewall [W. Fulp and Ryan J] consists of multiple firewalls coupled in parallel and a gatedevice, as shown in figure. Every firewall in the system outfits a local strategy, where incoming packets are disseminated across the firewalls for processing (one packet is sent to one firewall), permitting different packets to be processed in parallel. Because every packet is processed by the policy, reliability is maintained.

The common operation of the scheme can be described as follows. When a packet reaches to the function-parallel system it is promoted to all the firewall and the gate. Each firewall processes the packet with its local policy, as well as any state information. The firewall then signals the gate representing either no match was found, or offers the rule number and action if a match was found. As local policies are a division of the original, a no-matchis a suitable answer and is necessary for the function-parallel design. The gate stores the outcomes and establishes the final action to achieve on the packet using the Firewall rule.

Parallel Firewall System Components

A function parallel system consists of a collection of firewall nodes. Packets are replicated to all firewall nodes as they go into the system. Policies must be disseminated across the system such that they specify an accept set identical to the original accept set and no local policy's accept set can overlap another local policy's accept set. The received packets are then combined into one stream to arrive at the destination. A control plane is also essential to permit common system management.

Firewall Nodes

The firewall nodes contain a network interface card for every network input and control plane. This design implements the firewall nodes as PCs running the Linux operating system with a kernel that supports iptables. It must be stated that iptables was selected for an profitable feature of provisional rule dispensation. By default there are built-in sets of policies called chains divided by whether they hold traffic bound for procedures pay attention on local input devices, proposed for routing throughout the system to another network or outbound and supplied from a local procedure. In addition iptables has support for user defined chains. The user defined chains can be called if a packet matches a rule in a sequence.

Packet Duplicator

Packet Duplicator is essential on all links which will input traffic into the system. In Ethernet networks packet duplication is easily achieved with a network hub because any packets arriving on a hub port are copied to all other hub ports. But, in high speed networks hub technology is not presented. The only available devices in high speed networks to attain duplication are network taps. These devices are used for duplicating network traffic, usually used in intrusion detection systems that necessitate network monitoring.

Control Plane

In a single firewall system, protected customer communication can be offered through the prerequisite of physical existence. To generate a protected technique of management, communication to the collection of firewall nodes can be quarantined on a separate control plane. In the simplest form this entails a separate subnet which all firewall nodes survive on.

Component Integration

Joining these components into a functional design can be cut down into two network representations. The first utilizes simply one packet duplicator and can simply offer protection for traffic from one source. The second topology allows two networks to communicate bi-directionally through the system.

One-Directional System

This technique only thinks about packets traveling in one direction; for example, only packets are moving from source to destination. The collection of firewall nodes use the same IP address and MAC address and all will take the incoming network interface card in immoral mode. All but one firewall node will allow existing networking equipment outside of the system to cooperate with no alterations. The firewall node which does respond to user requirements can also be permitted to respond to ICMP ping requests in order to formulate the system additional compatible. For the outgoing network interface cards, any IP or MAC can be used as long as it permits communication with the destination.

Bi-directional System

The Bi-directional system considers packets moving in both directions; for example, packets traveling from source to destination and then back to source. Setup the source network interface cards in the similar method as the one directional scheme. The destination network interface cards are setup using a different single IP and MAC. All Network Interface Cards are brought up in immoral mode, and only one destination Network Interface Card is allowed to react to user requests. Since this design needs that both network duplicate packets, the tap device must be able to combine traffic which is send back through it onto the linkage with the destination network.

Dynamic Update of Firewall Policy

A firewall is a system component that pedals the traversal of packets crossways the limitations of a protected system support on an exact safety strategy. A firewall safety strategy is a list of planned sort out rules that define the events achieve on matching packets. A regulation is collected of sort out pasture such as protocol type, starting place IP address, end IP address, starting place port and end port, and an achievement pasture.

Every network pasture could be a on its own value or variety of standards. Filtering events are also to recognize, which exceed the packet into or from the protected system, or to Denny, which causes the container to be redundant. The containers are established or without by an exact regulation if the packet header in order matches all the system pasture of this rule. Without a firewall policy, overseer and connection are brief unsighted. Without a policy to guide firewall execution and direction, the firewall itself may become a safety problem [E. AI-Shaer and H. Hamed, 2002].

General work has been done to encourage the filtering efficiency of firewall. For a firewall, one of the assessment indicators, which make a decision its presentation, is tuple evaluation [E.W. Fulp and S. J. Tarsa, 2005]. When firewall strategy alters, like a regulation deleted, these implement will not energetically reply to the updating. Differing to that, they all require recreating the firewall strategy.

As the request of system develops, the firewall policies could change regularly. If every change leads to a renovation of the firewall, the price would be enormous. Although it is rather ancient, file explanation of firewall strategy is able to reply to the inform straightly. On the other side, some of the communicate tools are not capable to represent all the matching type of filtering field, like Policy Tree, this would limit the request of these tools. To pick up the filtering speed of firewall, researchers have planned many appearance tools for firewall policy. However, these tools divide a limitation: not well-matched with active updating of firewall strategy. Therefore, this paper recommend Neural Network model. NN can handle not only the package filtering but also energetically reply to the updating of innovative policies.

In this study, we examine the use of neural network for packet filtering. The neural network system was considered in eight ways with effort to the neural network in the form of moreover access rules or optimized permission rules or binary form of access rules or representing wildcards as 0 and 255, or grouping of them. These qualified neural networks were analyzing for their accuracy and the presentation aspects such as preparation time using test data. In order to additional recover the safety; the data connected to the local usage of the network was also used to prepare the network

Neural Network Model:

Neural network is one of the inter-disciplines increasing quickly in recent years. It is collected of a great deal of simple processing units (neuron) with flexibility, interconnecting as a neural network [Kobayashi, F.; Fukui, T.; Arai, F, 2002]. It can suggest the information dispensation functions of human brain, with great abilities of nonlinear approximation, information storage, large-scale parallel processing, self-training study, self-organization, fault-tolerant and so on.

The information dispensation in the neural network is understood by the contact between the neurons, and the storage of knowledge and in order is accessible as dispersed physical interconnection of the network components. The study and classification of the neural network depend on the energetic development of the association weights between the neurons.

Firewall Policy Editor

Firewall policies [E. AI-Shaer and H. Hamed, 2002] are often written by dissimilar network overseer and irregularly updated (by inserting, modifying or removing rules) to contain new security requirements and network topology changes. Editing a safety policy can be far harder than creating a new one. As rules in firewall policy are prepared, a new regulation must be inserting in a exacting order to keep away from produce anomalies. The same is valid if any network field in a rule is customized. In this section, we present policy editor tools that simplifies the rule editing task considerably, and avoids bring in irregularity due to policy modernize.

The policy editor (1) without delay the user with the proper position(s) for a new or customized rule (2) shows the changes in the safety policy semantic before and after take away a rule, and (3) give visual aids for users to track and confirm policy changes. Using the policy editor, overseer requires no preceding information or understating of the firewall policy in order to put in, adapt or take away a rule.

Rule Insertion

Since the arrange of rules in the filtering act list directly collision the semantics of the firewall safety policy, a new rule must be introduce in the correct order in the policy such that no investigation or idleness is formed. The policy editor helps the user to determine the correct position(s) of the new rule to be inserted. It also recognizes irregularity that may occur due to improper introduction of the new rule.

The general idea is that the order of a new rule is resolute based on its relation with other accessible system in the firewall strategy. In universal, a new regulation should be inserted before any rule that is its superset match, and after any rule that is its separation match. The strategy tree is used to keep track of the correct order of the new rule, and find out any potential anomalies. The algorithm execute the mechanism to insert a new rule is fully illustrate in [E. AI-Shaer and H. Hamed, 2002].

The algorithm is prepared into two stages: the browsing stage and the introduction stage. In the browsing stage, the fields of the new rule are comparing with the equivalent tree branch values one at a time. If the field value of the new regulation is a subset of an accessible branch, then the new rule must be inserting before the least amount order of all the system in this branch. If the field value is a superset of an existing branch, the rule must be inserted after the highest order of all the system in this branch. In addition, if the field value is an exact match or a subset match of a branch, appraise the next field keep on recursively by browsing through the division sub-tree awaiting correct position of the rule within the sub-tree is resolute. Otherwise, if displace or superset match happen, a division is created for the new rule.

The algorithm enters into the introduction stage when the achievement field of a new rule is to be inserted. If an action division is created for the new rule, then the rule will be inserted and allocate the order resolute in the browsing stage. If there is more than one possible order for this rule, the user is request to choose an order from within a valid range of instructions as strong-minded in the browsing stage. However, if the order state of the new rule remainder undecided then policy editor discards this new rule and rapid the user with the appropriate significance. If the rule is put in, the anomaly detection algorithm is invoked to prepare the overseer with any generalization or correlation cases as a potential cause of anomalies in the firewall strategy.

Rule Removal and Modification

In general, removing a rule has much less collision on the firewall strategy than introduction [E. AI-Shaer and H. Hamed, 2002]. An uninvolved rule does not bring in an irregularity but it might alter the strategy semantics and this change should be decorated and established. To remove a regulation, the user goes into the rule number to reduce the rule from the rule list and choose to take away it. To sample the effect of rule elimination, the policy editor gives a textual conversion of the exaggerated portion of the strategy before and after the rule is uninvolved. The user is able to contrast and examine the strategy semantics before and after elimination, and re-assure accuracy of the strategy changes. Modifying a rule in a firewall policy is also a dangerous process. However, this restriction action can be easily managed as rule taking away and insertion as explain before.


It is essential that a network firewall performs evidently to legal users, with small or no effect on the supposed network performance. This is particularly true if traffic necessitates specific network Quality of Service (QoS), such as limits on the packet delay, jitter, and throughput. Due to the increasing traffic loads and network speeds the firewall can rapidly turn into a blockage. The proposed parallel design consists of several firewalls and Neural Network for dynamically updating the Firewall Policy. Every firewall outfits a part of the defense policy. When a packet enter into the system it is processed by all firewall concurrently, which considerably decrease the processing time per packet. In addition, rule allocation guidelines that preserve policy reliability were introduced, which guarantees the parallel design and a conventional single firewall constantly arrive at the same conclusion for any packet. In addition unlike other designs, the proposed design can offer stateful assessments since a packet is routed by every firewall. Therefore, the function-parallel firewall design is a scalable solution that can offer superior performance and additional potentials than other designs.


Subrata Acharya,, Jia Wang, Zihui Ge, Taieb F. Znati,_ and Albert Greenberg “Traffic-Aware Firewall Optimization Strategies” 2006.

Toshiya Okabe Tsutomu Kitamura Takayuki Shizuno “Statistical Traffic Identification Method”,2006.

Based on Flow-Level Behavior for Fair VoIP service, 2006.

Thomas j. walsh and d. richard kuhn “Challenges in Securing Voice over IP “2007.

El-Sayed M. El-Alfy and Shokri Z. Selim “On Optimal Firewall Rule Ordering “,2007.

A Function-Parallel Architecture for High-Speed Firewalls Errin W. Fulp and Ryan J. Farley Department of Computer Science Wake Forest University, Winston-Salem, NC 27109-7311, USA Email: fulp@wfu.edu nsg.cs.wfu.edu

Carsten Benecke. A parallel packet screen for high speed networks. In Proceedings of the 15th Annual Computer Security Applications Conference, 1999.

Errin W. Fulp. Firewall architectures for high speed networks. Technical Report 20026, Wake Forest University Computer Science Department, 2002.

Kobayashi, F.; Fukui, T.; Arai, F, “Sensor selected fusion with sensor selection based gating neural network”, Fuzzy Systems, 2002. FUZZ-IEEE'02. Proceedings of the 2002 IEEE International Conference on Volume 2, 12-17 May 2002 Page(s):1482 - 1487.

E.W. Fulp and S. J. Tarsa. Trie-based policy representations for network firewalls. In Proc. of the 10th IEEE Symposium on Computers and Communications (ISCC 2005), pages 434– 441, June 2005.

E. AI-Shaer and H. Hamed. “Design and Implementation of Firewall Policy Advisor Tools.” Technical Report cTI-lechmpO801, School of Computer Science Telecommunications and lnformation Systems, DePaul University, August 2002