Abstract: today cyber-security becomesa need as it provides protection from highly vulnerable intrusions andthreats.it is impractical for human without considerable automation to handlecyber threat and highly vulnerable intrusions. To handle this situation, it needs to develop sophisticated, flexible,robust and adaptable software also called cyber defense system. This is enoughintelligent system to detect a variety of threats, refine and update thesetechnologies to combat it. Intrusion Detection system, Data Mining, and Computational Intelligence system areArtificial Techniques which provide detection and prevention of highly vulnerable threats and intrusions. Theaim of this paper to present the progress in the field of AI for defending fromcyber-crimes, to describe how these techniques are effective as well as providethe scope of future work.
Keywords: Artificial Intelligence, Data mining, Cyber Defense system, IntrusionDetection System, Computational Intelligence system1. INTRODUCTIONCybercrime is a most complex problemin the cyberworld.it is defined as any illegal activity that applied to a computer to harm the system or systemfiles and the computer security.A recent study on cybercrime shows that it is impractical to handlecyber-crimes for human without considerable automation. Furthermore, conventionally fixed algorithms are also notenough to handle the dynamically evolving cyber threats. To handle thissituation, it needs to developsophisticated and flexible software for protection and prevention from cyberthreats.
Cyber Defense system able to detect many of the cyber-attack andalerts the system. Human intervention is simply not enough to analysis the cyber threats and appropriateresponse. Cyber-attack is carried out with smart agents of worms and viruses.Smart semi-autonomous agents used to defend againstcyber-threats. This so-called system able to find out the type ofthreat, the response of threat, and the object of threat.
it also able to findout how to check and stop the secondaryattack. A variety of CDS were introduced but there is need to refine andupdate CDS to introduce the various techniques of AI. These techniques improve the security measures.Artificial intelligence offers manycomputing methods like Data Mining,Computational Intelligence System, Intrusion DetectionSystem, Neural Network, Pattern Recognition, Fuzzy Logic, Machine Learning,Expert System, Intelligent Agents, Search, Learning, Constraint Solving etc.Computational Intelligence System, DataMining, and Intrusion Detection Systemhave furthermore typed.
Data Mining technique is applied to observe the intrusions by recognizing the patterns of program and useractivity. .Association, Clustering, Classification, Prediction, and Sequence Patterns are Data Mining techniques.The Computational Intelligent System usually includes Fuzzy Logic, Evolutionary Computation, Cellular Automata,Intelligent Agent Systems, ANN, Artificial Immune Systems models. Thesetechniques allow efficient decision making. Theartificial immune system model is taken from the immune system. The BiologicalImmune System is natural defense system providesprotection against averse to many diseases.
Artificial Immune System,Artificial Neural Network, Genetic Algorithms are important techniques ofArtificial Immune System.Intrusion Detection (ID) is a process to monitor the traffic in thenetwork and monitor the strange activities and alert the system as well as a network administrator. Intrusion Prevention(IP) is the procedure of observing thetraffic in the network, used to identify the threats and respond it quickly.IDPS used to detect the problems in the network and solve these problems.
Herepresent three types of IDPS, first is network based and second host-based and thirdis a honeypot. There are 2 types of IDSanomaly and misuse detection.The second session of the present paperintroduces the existing techniques ofartificial intelligence in information technology security. The third session explains the existing techniques of data-miningin the information technology security.
The fourth session explains the computational intelligent systemin cybersecurity. The fifth session explainsthe existing techniques of IDS in cybersecurity. The Sixth session explains the abbreviation and acronyms and theseventh session explains the conclusion and future scope. Hence, in this paper, by implement AI on ICDS is proposed to make thedefense system more effective.
2. ARTIFICIALINTELLIGENCEAI is an electronic machine that is enough intelligentto act like human beings. It resolves thecomplicated problems rapidly than human beings such as playing the chess game. This paper represents the specific method of AI to solve cybercrimes. Thesemethods are described here.2.1.
Artificial Neural NetsArtificial Neural Net is introduced afterinspiring the Natural Biological Nervous System.A Neuron is formed by interconnected processing components. ANN consists of a number ofartificial neurons.it works like a human brainbut it has fewer complex neuronconnection than the biological nervous system.
Neuron received a lot of inputs and rapidly parallel respond to it. A neural net beginswith the invention of perceptron by Frank Rosenblatt in 1957.the main featureof ANN is rapidly responding and speed ofoperation. ANN is mainly configured for learning, classification, for recognizing the pattern.it is also applied toselect the appropriate response. An ANN is applied for DOS recognition in the network, wormrecognition in computer, malware recognition in the computer, and for zombie recognition in computer and malwareclassification in forensic investigation.
ANN is well liked for its high speed to perform an operation.it may beimplemented in hardware as well as software. If it is implemented in hardware thanit is used in the graphics processor. Alot of technologies of ANN is developed such as third generation neural nets. A distinguish feature of ANNthat it is used for intrusion detection system and perform high-speed operations.2.2.Intelligent AgentsIntelligent agents are computer-generatedeffects that show respond when an unexpected event occurs.
They exchange information with each other for motility and flexibility in the environment to make the IA technology more effectively to combat against cyber-attack. IAgive more information about the cyber-attack .it work on internet and giveinformation without our permission.Intelligent behavior of intelligent agent makes them more special reactiveness, understanding of associateagent communication language, reactivity (ability to create some alternativesand to act).they use for mobility, reflection ability and for planning ability.It is used against DDOS. Intelligent agents are cooperative agents thatgive efficient defense against DOS and DDOS attack.
‘Cyber police’ consist of intelligent agentsis developed after solving some commercial, industrial and legal problems. Itsupports the intelligent agent’s quality and communication but inaccessible to foes.A multi-agent tool is required foran entire operating system of cyberspace such as a neural network-basedintrusion detection and hybrid multi-agent techniques.
One distinguishes application ofintelligent agent is agent communication language.2.3.Expert SystemAn expert system is mostcommonly used AI tool.
This system is used to get inquiries from system orclients to discover the answers. It supports direct decision support. Such asit is used in finance, medical diagnose and cyberspace.An expert system is used for small as well large and complex problems like in hybridsystem.
The expert system consistsof large knowledge, it stores allinformation regarding a specific application. Expert system shell (ESS) is usedto support the adding of knowledge in knowledge base expert system, it can beextended with the program to cooperatethe client as well as another programthat may be utilized in the hybrid expert system.ESS is empty knowledge base.Hence, to make an expert system, first select an expert system shell,second it gets knowledge about and filling the knowledge base with knowledge.
The second step is more complex and time-consuming.An Expert system is used is cyber defense. Itdetermines the safety efforts and helps how to use ideally in resources thatare limited in quantity.
it is used in network intrusion detection which is knowledge base. In short, the expert system is used to convert the systemknowledge into programming language code. For example,CD expert system is used for security planning.
2.4.Search The method is applied toresolve the complicated problems where there no other methods are applicable. People used it constantly intheir everyday life without knowing it.
General algorithm of search is used tosearch the problem, some of it is able tocheck the problem and provide a solutionanother only estimate the troubles. If additional knowledge adds tothe search algorithm than drastically improve the search. Search is almost usedin every intelligible program and it increasesthe efficiency of the program. Many search application used in the AI programto search the problem, for example, dynamic programming is applied to detectthe optimized security problem, it is hiddenfrom the system, it is invisible in AI applications. Such as alpha-beta search, search on trees, minimumsearch, and random search and so on. The ??-search is developed to use forcomputer chess .
divide and conquer is used in complex problems especially in that application where choose the best action.It is used to estimate the minimum and maximum possibilities. This enablesignore many of the options and speeds up the search.
2.5.Learning Learning is an extending knowledge system byarranging or extending the knowledge base. This is a significant problem of theArtificial Intelligence that is under consideration. Machine learning consists of a computationalmethod to add new knowledge, new skills and an advanced way to keep and organize the existing knowledge. Learning method contains two types ofmethod i.
e. supervised learning and unsupervised learning. This is useful whenmultiple types of data are present. Thisis commonly used in cyber defense whereabundant data exists. Data Mining is specifically elaborate forunsupervised learning in artificial intelligence.
Unsupervised is utilitarianfor neural nets, in particular, of autonomous maps. Parallel algorithm method is a learningmethod that executes on hardware. Geneticalgorithms and ANNs help in representingthese strategies.
For example, Genetic algorithm and fuzzy logic are applied toobserve intrusions. In short, applications of learning are machine learning, supervised andunsupervised learning, malware detection, intrusion detection and for self-organized maps. Machine learning is enough intelligent system which is applied forpattern recognition.2.
6.Constraint Solving Constraint satisfaction method is applied inAI to discover solutions to thoseproblems that are introduced by a set of constraint on the solution e.g.logical statements, tables, equations, inequalities etc. A constraint solution is consist of a collection of tuples (ordered pair, row) thatmeet all restrictions. There are a lot of problems exist that have differentconstraint solution because solutiondepends on the character of constraints. Such as constraints on finite sets, functionalconstraints, rational trees etc. In abstract level, almost every problem isrepresented as a constraint solving problem.
Constraint satisfaction method isused in decision making and situation analysis in AI. TABLE (I): APPLICATION OF AI METHODS AI METHODS Applications ANN(Artificial Neural Nets) Defence against DDOS For Forensic investigation For intrusion detection Very high speed of reaction Worm detection Intelligent Agent Mobility Rapid response ACL Defence against DOS Reactive Expert system the knowledge base for decision making for intrusion detection and prevention Search for decision making for searching algorithm the knowledge base Learning for malware detection for intrusion detection for machine learning for supervised learning for autonomous maps Constraint solving for constraint problem for quick decision determining for situation examine 3. DATA MINING DataMining technique isapplied to observe the intrusions by recognizingthe patterns of program and user activity. Association, prediction, clustering,classification, and sequence patterns aredata mining techniques. 3.1.
Association Association rules in data mining are a conditional statement that exposes the connection among seeminglyunconnected figures and characters in RDBMS for example if a person buys a kg sugar, he is 75% likely to purchasemilk.3.2.Classification Classification in data mining is a method to assign a group of items to specifictarget classes. The function of this method is to estimate the aimed class foreach instance in the data. E.
g.A classificationmodel used to identify the vulnerabilities in the Nessus as low, medium, highand critical. Classification is separateand does not imply the order. It classifies the predefined data in multipleitems of the same quality.3.3.
Clustering Same quality ofobjects are in one class is called a cluster.A process to collect the same quality of data in a class is a cluster. The big benefit of the cluster methodis to distinguish between different groups and also objects of differentquality.3.4.Prediction Prediction is DataMining method which estimates a persistent value function and sequence valuefunction.it also predicts the relationship between dependent and independentvariables.
For example data analysis task in datamining.3.5.Sequential Patterns It is data mining technique to recognize statisticalrelevant patterns between data, such as consider a sequence database torepresent the client’s purchases from the general store. TABLE (II). FUNCTIONS OF DATA MINING TECHNIQUES DM Techniques Function Association Method that discovers the relationship between an item with respect to another Classification Method to classify the items into the classes and categories.
It is separate and do not imply in order It is used for mathematical techniques such as decision trees, linear programming, and statistics. Clustering Used to collect the same quality object in a group Prediction Predict the relationship between dependent and independent variables Predict the relationship between continuous and order value function Sequence Patterns Identify the similar pattern in data transaction after a specific time order 4. COMPUTATIONALINTELLIGENT SYSTEM The Computational intelligent system usually includes Fuzzy Logic, Evolutionary Computation, Intelligent AgentSystems, Neural Networks, Cellular Automata, Artificial Immune Systems models.These techniques allow efficient decision making. The artificial immune system model is taken from the immune system. Thebiological immune system is natural barricade system which produces defense-averseto many diseases.
Artificial neural network, genetic algorithms are importanttechniques of the artificial immune system(AIS) model.4.1.Artificial Immune SystemTheartificial immune system isinvented after inspired by the natural immune system.(HIS) the human immune system is natural defense system against diseases.it is very complex system andcomprises of many dendritic cells T cells, B cells. D cells gain theinformation about antigen and dead cells. T cells are built in bone marrowand remove infectious cells present in the blood.
B cells are white cell andproduce antibodies. Today the artificial immune system isused in intrusion detection system, system optimization and in dataclassification.it is also comprised of dendritic cells.
Nowadays, a newsecurity-crime interest cache poisoning (ICP) attack is introduced into the network layer which destroys the routing packets. Both dendriticcells and directed diffusion responsible for the detection of anomalous behaviorof the junction, also recognize theantigens. Direct diffusion responsible for two packets and two tablesconsequently interest packet and data packet, interest data, and cache data.
Artificial Immunesystem better the detection process as it detectsmany anomalies in a network such as DOS,DDOS, R2L, U2R and probing.it also detect the MAC layer gene and routing layersecurity attack. The architecture of IDS using AIS. Fig.1:Architecture of IDS using AIS 4.
2.Artificial Neural Nets Artificialneural nets are invented based on the human nervous system (HIS). HIS composedof neurons that are interconnected with each other.it is responsible for Defence againstDDOS, for forensic investigation, for intrusion recognition,high speed of appropriate respond and decision making. Fig.2: General Architecture ofneuron 4.3.Association Genetic algorithm(GA) is introduced based on human natural selection, evolutionary theory andmainly on genetic inheritance.
A geneticalgorithm is used to solve the complicated problems.it is responsiblefor robust, adaptive and optimal solutions for many complicated problems. Agenetic algorithm is used for intrusion detection in network security(NS).It is also applied for classification of security attack. Fig.
3: GeneralArchitecture of Genetic Algorithm TABLE (III). USES OF COMPUTATIONAL INTELLIGENCE SYSTEMAPPLICATION Computational intelligence system application Uses of Computational intelligence system application Artificial immune system Intrusion detection Data classification System optimization Detection of R2L, u2R MAC layer gene and routing layer genetic attack Artificial Neural Nets Defence against DDOS For Forensic investigation For intrusion detection Very high speed of reaction Worm detection Genetic Algorithm For optimal solution For adaptive and robust solution For intrusion recognition For classification of security attack 5. INTRUSIONDETECTION SYSTEMIntrusiondetection is the process of monitor the traffic in the network and monitor thestrange activities and alert the system as well as a network administrator. There are three groups of IDS first isnetwork based and second host-based andthird is a honeypot. There two types of IDS.
There are two typesof IDS. Anomaly and misuse detection.5.1.Network-based A system that recognizes the intrusionafter monitoring the traffic in the network devices. For example Networkinterface card (NIC). 5.2.
Host-based It monitors the files and processactivities that associate with a software environment related to a specifichost. For example, blocking IDS thatrelate the Host-based IDS with modifiedfirewall rules.5.3.Honeypot It is introduced to trap the intruder, it traces down the location of theintruder and gives a response to the attack .
it work on the networkbase sensor.TYPES OF IDSThere two types ofIDS anomaly and misuse detection5.4.Anomaly Detection It is the abnormal behavior of thesystem. For example system calls etc. 5.5.Misuse Detection The method to penetrate a system.
These penetrationsare signature and pattern. These penetrations are static and set of sequence ofaction. The system responds differently depending on the penetrations. 6. ABBREVIATION ANDACRONYMS A. (AI) abbreviate as Artificial Intelligence: AIis an electronic machine that is enough intelligent to behave like the humanbeings.B. (DM)abbreviate as Data mining: Data miningtechnique is applied to observe the intrusions by recognizing the patterns of program and user activity.
C. (CDS)abbreviate as Cyber Defense system: Cyber Defense system able to detect many ofthe cyber-attack and alerts the system.D. (IDS)abbreviate as Intrusion Detection System: Intrusion detection (ID) is theoperation of monitor the traffic in the network and monitor the strangeactivities and alert the system as well as a networkadministrator.E.
(CIS)abbreviate as Computational Intelligence system: CIS allows efficient decisionmaking.F. (ML)abbreviate as Machine learning: Learning is an extending knowledge system by arranging or extending theknowledge base.G. (ES)Expertsystem: An expert system is most commonlyused AI tool. This system is used to get inquiries from system or clients todiscover the answers.H. (IA)abbreviate as intelligent agents: Intelligent agents are computer generatedforces that show respond when an unexpected event occurs.
I. (AIS)abbreviate as an Artificial immune system:The artificial immune system is inventedafter inspired by the natural immune system.(HIS) the human immune system is natural defense system againstdiseases.J. (ANN)abbreviate as an artificial neural network: Artificial Neural Net is introduced byinspiring the natural biological nervous system.K.
(GA)abbreviate as Genetic algorithms: Genetic algorithm (GA) is introduced based on human natural selection,evolutionary theory and mainly on genetic inheritance. A genetic algorithm is used to solve the complicated problems.L. (IPS)abbreviate as intrusion prevention system: Intrusion prevention (IP) is theprocedure of observing the traffic in thenetwork, used to identify the threats and respond it quickly. 7. FUTURE WORK ANDCONCLUSIONIn this paper present the defense against sophistication attack.
Application of AI used to increase the efficiency of the cyber defense system. This application monitors thestrange activity in the network, worm detection in the computer and alerts the system and administrator that someunwanted things occur. Combine the use of the different techniques of AI, DM,IDPS, and Computational intelligent system in the security management system toimprove the security defense against security threats and intrusions. Some AIand DM techniques applied in the cyber defense system to remove the immediatecyber defense problems that require moreintelligent solutions that are present.
In the future,some more of the applications of AI can be used for decision making andfurthermore for the cyber defense system. 8. ACKNOWLEDGMENT Sadaf Safdar thanks, DR. SherazAhmad Malik and DR. AWAIS for their helping in writing the paper and alsospecial thanks, DR. Sheraz for reviewing my paper and encourage me to submitit. I thank my co-authors for their contribution.
Lastly special thanks to theinstitute GCUF which supported us.