System and method to detect malicious user behaviour on a virtual communal network

IN595075BActive Publication Date: 2026-07-10VIT AP UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
IN202341057601
Authority / Receiving Office
IN · IN
Patent Type
Patents
Current Assignee / Owner
Filing Date
2023-08-28
Publication Date
2026-07-10
Estimated Expiration
2043-08-28

AI Technical Summary

Technical Problem

Traditional methods for detecting malicious behavior on virtual communal networks are inadequate due to reliance on static rules, supervised learning, and inability to adapt to evolving tactics, leading to inaccurate identification and real-time detection challenges, as well as high computational resource requirements.

Method used

A hybrid approach integrating user behavior analytics with Deep Reinforcement Learning (DRL) algorithms, specifically Deep Q-Network (DQN) and Proximal Policy Optimization (PPO), to analyze user interactions, content sharing, and network connections, enabling real-time adaptation and reducing false positives and negatives.

Benefits of technology

This solution enhances the accuracy and efficiency of malicious behavior detection, adapts to evolving threats, and reduces computational complexity, providing a reliable and cost-effective system for real-time identification of malicious activities.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The present disclosure relates to a system and method to detect malicious user behaviour on a virtual communal network. The method (300) includes analysing user behaviour patterns from various dimensions of user interactions, content sharing, temporal patterns, and network connections (302). The method (300) also includes extracting significant features for analysis (304). The method (300) further includes identifying potential malicious behaviour by incorporating Deep Q-Network (DQN), deep reinforcement learning (DRL) and proximal policy optimisation (PPO) techniques (306). The method (300) also includes utilizing learned optimum policies from DQN and PPO techniques (308).
Need to check novelty before this filing date? Find Prior Art

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of network security. Moreparticularly the present disclosure relates to system and method to detectmalicious behaviour on a virtual communal network.BACKGROUND

[0002] User security, confidentiality, and trust are vulnerable to malicious userson virtual communal networks. Traditional methods relying on static rules andsupervised learning often fall short in accurately identifying and mitigatingmalicious behaviour. This invention addresses these limitations by introducing ahybrid approach that integrates user behaviour analytics and advanced DeepReinforcement Learning (DRL) algorithms, such as Deep Q-Network (DQN) andProximal Policy Optimisation (PPO), to enhance the accuracy and efficiency ofmalicious behaviour detection.

[0003] Traditional methods often rely on static rules and predefined patterns todetect malicious behaviour. These methods struggle to adapt to new and evolvingtactics employed by malicious users, making them susceptible to evasionstrategies. Also, many conventional detection methods heavily depend onsupervised learning, where models are trained on labelled data. This approachrequires extensive labelled datasets that accurately represent the diversity ofpotential malicious behaviours. However, obtaining such data can be challenging,leading to biased or incomplete models.

[0004] In addition, traditional methods may lack the ability to understand thecontext in which user behaviours occur. They may not consider dynamic shifts inuser interactions, content sharing patterns, and evolving network connections,leading to false positives or false negatives. Further, user behaviours on virtualcommunal networks often exhibit long-term dependencies, where current actionsare influenced by past actions. Traditional methods that don't account for thesedependencies may miss identifying subtle patterns indicative of malicious intent.

[0005] Also, rapid changes in user behaviour and network dynamics requirereal-time adaptation of detection methods. Traditional techniques that lackdynamic learning and adaptation mechanisms may lag behind in identifying newthreats. Further, the conventional methods that process and analyse data usingtraditional machine learning algorithms might require substantial computationalresources, causing delays in detecting malicious behaviour.

[0006] Hence there is, therefore, a need of an improved system and method todetect malicious behaviour on a virtual communal network, to address theaforementioned issue(s).OBJECTS OF THE PRESENT DISCLOSURE

[0007] Some of the objects of the present disclosure, which at least oneembodiment herein satisfies are as listed herein below.

[0008] It is an object of the present disclosure to provide a system to detectmalicious behaviour on a virtual communal network.

[0009] It is another object of the present disclosure to provide a system whichaims to significantly improve the accuracy of detecting malicious behaviour onvirtual communal networks.

[00010] It is another object of the present disclosure to provide a system mayseek to enable real-time adaptation to evolving malicious tactics and patterns.

[00011] It is another object of the present disclosure to provide a system aims tocomprehensively analyse intricate user behaviour patterns across multipledimensions, including user interactions, content sharing, temporal patterns, andnetwork connections.

[00012] It is another object of the present disclosure to provide a system intendsto address the challenge of long-term dependencies in user behaviour byleveraging the capabilities of the Deep Q-Network (DQN) algorithm.

[00013] It is another object of the present disclosure to provide a system seeksto reduce both false positive and false negative rates.

[00014] It is another object of the present disclosure to provide a system whichdecrease complexity and is cost effective.

[00015] It is another object of the present disclosure to provide a system whichis power efficient.

[00016] It is another object of the present disclosure to provide a system whichis simple and easy to use, thereby being more reliable.SUMMARY

[00017] This summary is provided to introduce simplified concepts of a systemand method to detect malicious behaviour on a virtual communal network. Thissummary is not intended to identify key or essential features of the claimedsubject matter, nor is it intended for use in determining / limiting the scope of theclaimed subject matter.

[00018] The present disclosure generally relates to the field of network security.More particularly the present disclosure relates to system and method to detectmalicious user behaviour on a virtual communal network.

[00019] An aspect of the present disclosure relates to a system to detectmalicious user behaviour on a virtual communal network. The system includes aprocessor. The system also includes a memory comprising a set of instructions,which when executed by the processor cause the processor to: analyse userbehaviour patterns from various dimensions of user interactions, content sharing,temporal patterns, and network connections; extract significant features foranalysis; identify potential malicious behaviour by incorporating Deep Q-Network(DQN), deep reinforcement learning (DRL) and proximal policy optimisation(PPO) techniques; and utilizes learned optimum policies from DQN and PPOtechniques.

[00020] Another aspect of the present disclosure relates to a method fordetecting malicious user behaviour on a virtual communal network. The methodincludes analysing user behaviour patterns from various dimensions of userinteractions, content sharing, temporal patterns, and network connections. Themethod also includes extracting significant features for analysis. The methodfurther includes identifying potential malicious behaviour by incorporating DeepQ-Network (DQN), deep reinforcement learning (DRL) and proximal policyoptimisation (PPO) techniques. The method also includes utilizing learnedoptimum policies from DQN and PPO techniques.

[00021] Various objects, features, aspects and advantages of the presentdisclosure will become more apparent from the following detailed description ofpreferred embodiments, along with the accompanying drawing figures in whichlike numerals represent like features.

[00022] Within the scope of this application, it is expressly envisaged that thevarious aspects, embodiments, examples and alternatives set out in the precedingparagraphs, in the claims and / or in the following description and drawings, and inparticular the individual features thereof, may be taken independently or in anycombination. Features described in connection with one embodiment areapplicable to all embodiments, unless such features are incompatible.BRIEF DESCRIPTION OF THE DRAWINGS

[00023] In the figures, similar components and / or features may have the samereference label. Further, various components of the same type may bedistinguished by following the reference label with a second label thatdistinguishes among the similar components. If only the first reference label isused in the specification, the description is applicable to any one of the similarcomponents having the same first reference label irrespective of the secondreference label.

[00024] FIG. 1 illustrates an exemplary representation of a block diagramrepresenting a system to detect malicious behaviour on a virtual communalnetwork, in accordance with an embodiment of the present disclosure.

[00025] FIG. 2 illustrates an exemplary representation of the system or a server,in accordance with an embodiment of the present disclosure.

[00026] FIG. 3 illustrates a flow diagram illustrating a method for detectingmalicious behaviour on a virtual communal network, in accordance with anembodiment of the present disclosure.

[00027] FIG. 4 illustrates an exemplary computer system to implement theproposed system in accordance with embodiments of the present disclosure.DETAILED DESCRIPTION

[00028] The following is a detailed description of embodiments of thedisclosure depicted in the accompanying drawings. The embodiments are in suchdetail as to clearly communicate the disclosure. However, the amount of detailoffered is not intended to limit the anticipated variations of embodiments; on thecontrary, the intention is to cover all modifications, equivalents, and alternativesfalling within the scope of the present disclosure as defined by the appendedclaims.

[00029] In the following description, numerous specific details are set forth inorder to provide a thorough understanding of embodiments of the presentinvention. It will be apparent to one skilled in the art that embodiments of thepresent invention may be practiced without some of these specific details.

[00030] The present disclosure relates to the field of network security. Moreparticularly the present disclosure relates to system and method to detectmalicious behaviour on a virtual communal network.

[00031] FIG. 1 illustrates an exemplary representation of a block diagramrepresenting a system for edge-enabled malicious vehicle detection, in accordancewith an embodiment of the present disclosure. The system 100 includes aprocessor 102 operatively coupled to the distributed ledger. The system 100 alsoincludes a memory 104 comprising a set of instructions, which when executed bythe processor cause the processor 102 to execute certain process via a plurality ofmodules.

[00032] The system 100 may include an analysis module 106 which may beconfigured to analyse user behaviour patterns from various dimensions of userinteractions, content sharing, temporal patterns, and network connections.

[00033] The system 100 may also include an extraction module 108 which maybe configured to extract significant features for analysis.

[00034] Furthermore, the system 100 may include a malicious behaviouridentification module 110 which may be configured to identify potential maliciousbehaviour by incorporating Deep Q-Network (DQN), deep reinforcement learning(DRL) and proximal policy optimisation (PPO) techniques.

[00035] The system 100 may further include a utilization module 112 whichmay be configured to utilize learned optimum policies from DQN and PPOtechniques.

[00036] In one exemplary embodiment, the processor 102 may be configured toimplement Proximal Policy Optimisation (PPO) techniques for enhanceddetection efficiency. The processor 102 may also be configured to respond tochanges in malicious user behaviour by using adaptive policy adjustmentmechanism.

[00037] In one specific embodiment, the deep reinforcement learning (DRL)technique may employ deep neural networks and reinforcement learning conceptsto analyse and decipher user behaviour patterns.

[00038] In operation, user security, confidentiality, and conviction are extremelysusceptible by the persistent usage of malicious users on virtual communalnetworks. Due to their reliance on static rules and supervised learning approaches,traditional methods of detecting malicious behaviour recurrently fall short insuccessfully recognising and mitigating these risks. The system analyses a hybridtechnique for detecting malicious users on online social networks by assimilationof user behaviour analytics and Deep Reinforcement Learning (DRL) algorithms,predominantly Deep Q-Network (DQN) and Proximal Policy Optimisation (PPO).Automated systems are assumed the aptitude to analyse and decipher patterns ofuser behaviour by utilising DRL algorithms, which conglomerate deep neuralnetworks with reinforcement learning concepts. With higher precision and agilityin the dynamic and changing environment of virtual social networks, thisfacilitates the detection of potentially hazardous user behaviours. Further, thefundamental approach for identifying malevolent users is user behaviouranalytics. By extracting significant features that encapsulate the complexities ofuser behaviour by examining numerous dimensions of user interactions, contentsharing, temporal patterns, and network connections. These physiognomies act asinputs to the DRL models, enabling them to recognise intricate patterns anddetermine whether malicious behaviour is existent.

[00039] The DQN technique renders a robust basis for training agents toeffectively analyse and categorise human behaviour since it can handle high-dimensional input spaces and long-term dependencies. The DQN agent learnsoptimum policies that direct its decision-making process and enhance detectionperformance over time by utilising Q-learning and experience replaymethodologies. In order to enhance the system's detecting abilities, ProximalPolicy Optimisation (PPO), a cutting-edge policy optimisation technique isproposed. By using replacement objective functions and implementing policychanges vigilantly, PPO delivers increased stability and sampling efficiency. Thesystem can further improve its policies using PPO in order to respond to dynamicshifts in destructive user behaviour.

[00040] The system strengthens the competency of the DRL and PPOalgorithms for distinguishing detrimental malicious users on virtual communalnetworks and progresses the field of user behaviour analytics. Aim is to enhanceuser security, safeguard privacy, and promote a trustworthy and secureenvironment for social interactions by utilising cutting-edge technology. Theconsequences of this research exposed the door for the creation of intelligent,adaptable systems that can defend against new hazards in virtual communalnetworks.

[00041] Merging the strength of deep neural networks and reinforcementlearning algorithms, Deep Reinforcement Learning (DRL) approaches havedetermined substantial potential in contemporary era for elucidating complicateddecision-making issues. DRL lets an agent to learn from interactions with anenvironment, permitting it to maximize a reward signal by making sequentialdecisions based on the observed states. By utilizing DRL techniques, we may beable to improve the detection of harmful user behaviour by identifying complexpatterns and dynamics not considered by conventional methods.

[00042] The system emphases on Deep Q-Network (DQN) and Proximal PolicyOptimization (PPO), two popular DRL approaches. DQN is a model-free DRLtechnique that uses a deep neural network to approximate the Q-value function,countenancing it to activate with high-dimensional input data and recognizeintricate patterns. While maintaining stability during the learning process, PPO isa policy optimization technique that directly optimizes the policy parameters tomaximize the predicted reward.

[00043] Further, the prerequisite is to study how DQN and PPO possibly used toidentify malicious user behaviour. The foremost persistence is to capture theunderlying patterns and correlations in user behaviour that distinguish maliciousfrom benign actions by training a DQN model to interact with the systemenvironment and make sequential judgements. In order to improve the detectionsystem's performance even further, the integration of PPO with an emphasis onoptimizing the policy parameters and enhancing flexibility to shifting assaulttactics.

[00044] FIG. 2 illustrates an exemplary representation of the system or a server,in accordance with an embodiment of the present disclosure.

[00045] As illustrated, a module diagram 200 for the system 100 can maycomprise one or more processor(s) 202. It should be noted that the system 100 issubstantially similar to the system 100 of FIG. 1. The one or more processor(s)202 can be implemented as one or more microprocessors, microcomputers,microcontrollers, digital signal processors, central processing units, logiccircuitries, and / or any devices that manipulate data based on operationalinstructions. Among other capabilities, the one or more processor(s) 202 areconfigured to fetch and execute computer-readable instructions stored in amemory 204 of the system 100. The memory 204 can store one or more computer-readable instructions or routines, which can be fetched and executed to create orshare the data units over a network service. The memory 204 can comprise anynon-transitory storage device including, for example, volatile memory such asRAM, or non-volatile memory such as EPROM, flash memory, and the like.

[00046] The system 100 can also comprise an interface(s) 206. The interface(s)206 can comprise a variety of interfaces, for example, interfaces for data input andoutput devices, referred to as I / O devices, storage devices, and the like. Theinterface(s) 206 can facilitate communication of system 100. The interface(s) 206can also provide a communication pathway for one or more components of thesystem 100. Examples of such components include, but are not limited to,processing engine(s) 208 and data receiving engine 210, analysis engine 212. Itshould be noted that the processing engine 208 is substantially similar to theprocessor 102 of FIG. 1.

[00047] The processing engine(s) 208 can be implemented as a combination ofhardware and programming (for example, programmable instructions) toimplement one or more functionalities of the processing engine(s) 208. Inexamples described herein, such combinations of hardware and programming canbe implemented in several different ways. For example, the programming for theprocessing engine(s) 208 can be processor executable instructions stored on anon-transitory machine-readable storage medium and the hardware for theprocessing engine(s) 208 can comprise a processing resource (for example, one ormore processors), to execute such instructions. In the present examples, themachine-readable storage medium can store instructions that, when executed bythe processing resource, implement the processing engine(s) 208. In suchexamples, the system 100 can comprise the machine-readable storage mediumstoring the instructions and the processing resource to execute the instructions, orthe machine-readable storage medium can be separate but accessible to system100 and the processing resource. In other examples, the processing engine(s) 208can be implemented by electronic circuitry.

[00048] The data receiving engine 210 can comprise data that is either stored orgenerated as a result of functionalities implemented by any of the components ofthe processing engine(s) 208 or the system 100.

[00049] In an embodiment, the system 100 can include an analysis module 106which may be configured to analyse user behaviour patterns from variousdimensions of user interactions, content sharing, temporal patterns, and networkconnections.

[00050] In another embodiment, the system 100 can include an extractionmodule 108 which may be configured to extract significant features for analysis.

[00051] In another embodiment, the system 100 can include a maliciousbehaviour identification module 110 which may be configured to identifypotential malicious behaviour by incorporating Deep Q-Network (DQN), deepreinforcement learning (DRL) and proximal policy optimisation (PPO)techniques.

[00052] In another embodiment, the system 100 can include a utilization module112 which may be configured to utilize learned optimum policies from DQN andPPO techniques.

[00053] FIG. 3 illustrates a flow diagram illustrating a method for detectingmalicious behaviour on a virtual communal network, in accordance with anembodiment of the present disclosure.

[00054] As illustrated, in step 302, the method 300 includes analysing userbehaviour patterns from various dimensions of user interactions, content sharing,temporal patterns, and network connections.

[00055] As illustrated, in step 304, the method 300 includes extractingsignificant features for analysis.

[00056] As illustrated, in step 306, the method 300 includes identifyingpotential malicious behaviour by incorporating Deep Q-Network (DQN), deepreinforcement learning (DRL) and proximal policy optimisation (PPO)techniques.

[00057] As illustrated, in step 308, the method 300 includes utilizing learnedoptimum policies from DQN and PPO techniques.

[00058] In one embodiment, the user behaviour patterns are extracted frommultiple dimensions of user interactions, content sharing, temporal patterns, andnetwork connections.

[00059] In another embodiment, wherein the DRL technique comprises Deep Q-Network (DQN) and Proximal Policy Optimisation (PPO) techniques.

[00060] In one specific embodiment, the method may include employing DRLtechniques that integrate deep neural networks with reinforcement learningconcepts. In another embodiment, the method 300 may further include detectingpotentially hazardous user behaviours with higher precision and agility indynamic virtual social networks.

[00061] In one exemplary embodiment, the user behaviour analytics mayinvolve extracting significant features that encapsulate complexities of userbehaviour for analysis. In another embodiment, the significant features act asinputs to the DRL models, enabling recognition of intricate patterns anddetermination of malicious behaviour.

[00062] In one specific embodiment, the method 300 may further includeimplementing Proximal Policy Optimisation (PPO) for enhancing the system'sability for detecting malicious behaviour by utilizing replacement objectivefunctions and implementing vigilant policy changes. The method may furtherinclude utilizing PPO for adapting the system's policies to dynamic shifts indestructive user behaviour.

[00063] FIG. 4 illustrates an exemplary computer system to implement theproposed system in accordance with embodiments of the present disclosure.

[00064] As illustrated in Fig. 4, a computer system 400 can include an externalstorage device 410, a bus 420, a main memory 430, a read only memory 440, amass storage device 450, communication port 460, and a processor 470. A personskilled in the art will appreciate that the computer system may include more thanone processor and communication ports. Examples of processor 470 include, butare not limited to, an Intel Itanium or Itanium 2 processor(s), or AMDOpteron or Athlon MP processor(s), Motorola lines of processors,FortiSOC system on chip processors or other future processors. Processor 470may include various modules associated with embodiments of the presentinvention. Communication port 460 can be any of an RS-232 port for use with amodem based dialup connection, a 10 / 100 Ethernet port, a Gigabit or 10 Gigabitport using copper or fiber, a serial port, a parallel port, or other existing or futureports. Communication port 460 may be chosen depending on a network, such aLocal Area Network (LAN), Wide Area Network (WAN), or any network towhich computer system connects.

[00065] Memory 430 can be Random Access Memory (RAM), or any otherdynamic storage device commonly known in the art. Read-only memory 440 canbe any static storage device(s) e.g., but not limited to, a Programmable Read OnlyMemory (PROM) chips for storing static information e.g., start-up or BIOSinstructions for processor 470. Mass storage 450 may be any current or futuremass storage solution, which can be used to store information and / or instructions.Exemplary mass storage solutions include, but are not limited to, ParallelAdvanced Technology Attachment (PATA) or Serial Advanced TechnologyAttachment (SATA) hard disk drives or solid-state drives (internal or external,e.g., having Universal Serial Bus (USB) and / or Firewire interfaces), e.g. thoseavailable from Seagate (e.g., the Seagate Barracuda 7102 family) or Hitachi (e.g.,the Hitachi Deskstar 7K1000), one or more optical discs, Redundant Array ofIndependent Disks (RAID) storage, e.g. an array of disks (e.g., SATA arrays),available from various vendors including Dot Hill Systems Corp., LaCie, NexsanTechnologies, Inc. and Enhance Technology, Inc.

[00066] Bus 420 communicatively couple processor(s) 470 with the othermemory, storage and communication blocks. Bus 420 can be, e.g. a PeripheralComponent Interconnect (PCI) / PCI Extended (PCI-X) bus, Small ComputerSystem Interface (SCSI), USB or the like, for connecting expansion cards, drivesand other subsystems as well as other buses, such a front side bus (FSB), whichconnects processor 470 to software system.

[00067] Optionally, operator and administrative interfaces, e.g. a display,keyboard, and a cursor control device, may also be coupled to bus 420 to supportdirect operator interaction with a computer system. Other operator andadministrative interfaces can be provided through network connections connectedthrough communication port 460. The external storage device 410 can be any kindof external hard-drives, floppy drives, IOMEGA Zip Drives, Compact Disc -Read Only Memory (CD-ROM), Compact Disc-Re-Writable (CD-RW), DigitalVideo Disk-Read Only Memory (DVD-ROM). Components described above aremeant only to exemplify various possibilities. In no way should theaforementioned exemplary computer system limit the scope of the presentdisclosure.

[00068] Moreover, in interpreting the specification, all terms should beinterpreted in the broadest possible manner consistent with the context. Inparticular, the terms "comprises" and "comprising" should be interpreted asreferring to elements, components, or steps in a non-exclusive manner, indicatingthat the referenced elements, components, or steps may be present, or utilized, orcombined with other elements, components, or steps that are not expresslyreferenced. Where the specification claims refer to at least one of somethingselected from the group consisting of A, B, C ….and N, the text should beinterpreted as requiring only one element from the group, not A plus N, or B plusN, etc.

[00069] While the foregoing describes various embodiments of the invention,other and further embodiments of the invention may be devised without departingfrom the basic scope thereof. The scope of the invention is determined by theclaims that follow. The invention is not limited to the described embodiments,versions or examples, which are included to enable a person having ordinary skillin the art to make and use the invention when combined with information andknowledge available to the person having ordinary skill in the art.ADVANTAGES OF THE INVENTION

[00070] The proposed invention provides a system to detect maliciousbehaviour on a virtual communal network.

[00071] The proposed invention also provides a system which aims tosignificantly improve the accuracy of detecting malicious behaviour on virtualcommunal networks.

[00072] The proposed invention also provides a system may seek to enable real-time adaptation to evolving malicious tactics and patterns.

[00073] The proposed invention also provides a system aims tocomprehensively analyse intricate user behaviour patterns across multipledimensions, including user interactions, content sharing, temporal patterns, andnetwork connections.

[00074] The proposed invention also provides a system intends to address thechallenge of long-term dependencies in user behaviour by leveraging thecapabilities of the Deep Q-Network (DQN) algorithm.

[00075] The proposed invention also provides a system seeks to reduce bothfalse positive and false negative rates.

[00076] The proposed invention also provides a system which decreasecomplexity and is cost effective.

[00077] The proposed invention also provides a system which is powerefficient.

[00078] The proposed invention also provides a system which is simple andeasy to use, thereby being more reliable.

Claims

1. A system (100) to detect malicious user behaviour on a virtual communal network, wherein the said system (100) comprises of: a processor (102); a memory (104) comprising a set of instructions, which when executed by the processor (102) cause the processor to: analyse user behaviour patterns from various dimensions of user interactions, content sharing, temporal patterns, and network connections; extract significant features for analysis; identify potential malicious behaviour by incorporating Deep QNetwork (DQN), deep reinforcement learning (DRL) and proximal policy optimisation (PPO) techniques; and utilize learned optimum policies from DQN and PPO techniques.

2. The system (100) as claimed in claim 1, wherein the processor (102) is configured to: implement Proximal Policy Optimisation (PPO) techniques for enhanced detection efficiency; and respond to changes in malicious user behaviour by using adaptive policy adjustment mechanism.

3. The system (100) as claimed in claim 1, wherein the deep reinforcement learning (DRL) technique employs deep neural networks and reinforcement learning concepts to analyse and decipher user behaviour patterns.

4. A method (300) for detecting malicious user behaviour on a virtual communal network, wherein the said system (100) comprises of: analysing user behaviour patterns from various dimensions of user interactions, content sharing, temporal patterns, and network connections; (302) extracting significant features for analysis; (304) identifying potential malicious behaviour by incorporating Deep Q- Network (DQN), deep reinforcement learning (DRL) and proximal policy optimisation (PPO) techniques; and (306) utilizing learned optimum policies from DQN and PPO techniques. (308).

5. The method (300) as claimed in claim 4, wherein the user behaviour patterns are extracted from multiple dimensions of user interactions, content sharing, temporal patterns, and network connections.

6. The method (300) as claimed in claim 4, wherein the DRL technique comprises Deep Q-Network (DQN) and Proximal Policy Optimisation (PPO) techniques.

7. The method (300) as claimed in claim 4, comprising: employing DRL techniques that integrate deep neural networks with reinforcement learning concepts; and detecting potentially hazardous user behaviours with higher precision and agility in dynamic virtual social networks.

8. The method (300) as claimed in claim 4, wherein user behaviour analytics involves extracting significant features that encapsulate complexities of user behaviour for analysis.

9. The method (300) as claimed in claim 4, wherein the significant features act as inputs to the DRL models, enabling recognition of intricate patterns and determination of malicious behaviour.

10. The method (300) as claimed in claim 4, comprising: implementing Proximal Policy Optimisation (PPO) for enhancing the system's ability for detecting malicious behaviour by utilizing replacement objective functions and implementing vigilant policy changes; and utilizing PPO for adapting the system's policies to dynamic shifts in destructive user behaviour.