Women Safety Analytics – Protecting Women From Threats

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 02 | Feb 2026 www.irjet.net p-ISSN: 2395-0072

Women Safety Analytics – Protecting Women From Threats

Mrs.Aruna.M1,Arthi.B2 , Charan Teja.G3 , Ratnakar.K4 , Sai Karthik.K5

1Assistant Professor, Department of IT, TKR College of Engineering and Technology, Telangana, India 2,3,4,5B.Tech Students, Department of IT, TKR College of Engineering and Technology, Telangana, India ***

Abstract - With the rapidgrowth of socialmediaplatforms, online harassment, cyberbullying, and other forms of digital abuse have become significant threats to user safety, particularly for women. Detecting and preventing such harmful interactions requires advanced automated systems capable of understanding context and intent in textual data. This projectpresentsacomprehensiveWomenSafetyAnalytics system that leverages the Gemini 2.5 Flash AI model through Lang Chain for automated comment classification and risk assessment. The system collects user-generated comments, analyzes them in real time, and categorizes them into predefined threat types such as cyberbullying, threats and intimidation, sexual harassment, hate speech, cyberstalking, and human trafficking indicators. Comments identified as harmful trigger automated email notifications to users, ensuring timely awareness and preventive action. An admin dashboard enables administrators to monitor user activity, review comment analysis results, andmanage user activation efficiently. The proposed system demonstrates high accuracy, scalability, and flexibility in detecting abusive behavior while maintaining user privacy. The solution contributestocreating safer online environments and promoting digital safety and user well-being.

Key Words: Women Safety, Safety Analytics, Threat Detection, Real-Time Monitoring, Emergency Alert System, Machine Learning, Artificial Intelligence, Crime Prevention, Predictive Analytics, Location Tracking, IoT Sensors, Smart Surveillance, Mobile Safety Applications, Risk Assessment, Public Safety Systems.

1. INTRODUCTION

SocialmediaplatformssuchasFacebook,Twitter,Instagram, and WhatsApp have become essential for communication, learning,business,andemploymentopportunities.However, theincreasinguseoftheseplatformshasalsoresultedina rapid growth of online threats such as cyberbullying, harassment,hatespeech,intimidation,andonlinestalking. Women are often the major victims of such abuse, which causes emotional stress, fear, anxiety, and long-term psychological harm. Since online content is produced continuouslyatalargescale,traditionalmanualmonitoring and reporting mechanisms are not sufficient for effective protection.

1.1 Background and Motivation

Cyberbullyingandonlineharassmenthavebecomeserious issues in modern digital society. Several studies have exploredtheuseofmachinelearninganddeeplearningto detect cyberbullying content on social media. Research showsthatdeeplearning-basedclassifiersprovideimproved accuracy comparedto traditional classifierssuchasNaïve BayesandDecisionTrees,especiallywhenhandlinginformal language,slang,andevolvingabusivepatterns[9].Moreover, manyonlineharassmentincidentsarerepeatedovertime, which highlights the need for automated and scalable monitoringsolutions[10].

1.2 Limitations of Existing Approaches

Althoughexistingresearchprovideseffectivecyberbullying and hate speech detection methods, many systems still struggle with subtle and contextual abuse. Hate speech detection models often misclassify sarcasm or indirect insults due to the complexity of human language [10]. Traditionalkeyword-basedsystemsalsofailtounderstand semanticmeaningandcontext.Studieshighlightthatword embeddingsandNLP-basedcontextualunderstandingsare essentialforimprovingclassificationperformance[5],[6].In addition, many existing systems lack real-time alert mechanismsandintegrateddashboardsforadministrators.

1.3 Proposed Research Direction

To overcome these limitations, this research proposes a WomenSafetyAnalyticssystemthatintegratesAI-basedtext classificationusingNaturalLanguageProcessing.Thesystem detects multiple types of threats including cyberbullying, sexualharassment,threats,hatespeech,cyberstalking,spam, and human trafficking indicators. It automatically stores analysisresultsinadatabase,sendsemailalertstousersfor harmful content, and provides a centralized admin dashboard for monitoring and user management. This approach aims to improve online safety for women by enabling proactive detection, timely alerts, and scalable moderation.

2. PROPOSED SYSTEM

TheproposedWomenSafetyAnalyticssystemisdesignedto provide a comprehensive solution for detecting and managing online threats targeting women. The system integratestheGemini2.5FlashAImodelwithLangChainto

International

Volume: 13 Issue: 02 | Feb 2026 www.irjet.net p-ISSN: 2395-0072

analyzeuser-generatedcommentsinrealtime.Usersregister by providing basic details such as name, email, mobile number, and profile image. Accounts remain inactive until approved by the administrator. Submitted comments are processedbytheAImodelandclassifiedintocategoriessuch as cyberbullying, threats, sexual harassment, hate speech, cyberstalking, subtle abuse, spam, human trafficking indicators,orsafecontent.Whenharmfulcontentisdetected, automated email notifications are sent to users. An admin dashboard enables monitoring of user activity, comment analysis,andaccountmanagement.Alldataissecurelystored toensureprivacyandreliability.Thesystemisscalableand capableofhandlinglargevolumesofdataefficiently.

2.1 System Architecture

This diagram illustrates a three-tier web application architecture that shows how users and administrators interactwithasystemthroughclearlyseparatedlayers.Both the Userand Adminaccess the system via a web browser, which connects to the Presentation Layer (Web Interface) responsiblefordisplayingpagesandcollectinginputs.The presentation layer forwards user requests to the Business Logic Layer (Django Views & Controllers), where the core application logic is processed, such as handling rules, validations, and decision-making. This layer then communicateswiththeDataLayer(Database&Models)to performqueriesandupdatesonstoreddata.BoththeUser and Admin access the system via a web browser, which connects to the Presentation Layer (Web Interface) responsiblefordisplayingpagesandcollectinginputs.Both the Userand Adminaccess the system via a web browser, which connects to the Presentation Layer (Web Interface) responsiblefordisplayingpagesandcollectinginputs.Both the Userand Adminaccess the system via a web browser, which connects to the Presentation Layer (Web Interface) responsiblefordisplayingpagesandcollectinginputs. After processing, responses flow back upward from the data layer to the business logic layer, and finally to the presentationlayer ensuringacleanseparationofconcerns, betterscalability,andeasiermaintenanceoftheapplication.

2.2 Web-Based Userand Admin Interaction Module

Theweb-basedinteractionmodule formsthepresentation layer of the proposed system, enabling both users and administrators to access the platform through a browserbasedinterface.Userscanregister,login,sharetheirlocation, andtriggeremergencyalertswhentheysensedanger,while administratorscanmonitoractivities,manageuserdata,and analyze reported incidents. This layer focuses on usability andresponsiveness,ensuringthatcriticalsafetyfeaturesare easily accessible during emergencies. By acting as an interfacebetweenendusersandthebackendservices,the presentationlayerensuressmoothcommunicationandrealtimeresponsedelivery.

2.3 Backend Processing and Secure Data Management

Thebackendprocessingmodulerepresentsthebusinesslogic anddatalayersofthesystem,whereallcorefunctionalities are executed. The business logic layer handles request validation, threat analysis, alert generation, and communication with emergency services using predefined rules and analytics models. The data layer securely stores user profiles, location data, incident logs, and system configurationsinastructureddatabase.Thislayeredbackend architectureimprovessystemreliability,datasecurity,and scalability, making it suitable for real-time women safety analyticsandlong-termcrimeanalysis.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 02 | Feb 2026 www.irjet.net p-ISSN: 2395-0072

3. IMPLEMENTATION DETAILS

The Women Safety Analytics system is implemented as a web-basedapplicationthatintegratesArtificialIntelligence andNaturalLanguageProcessingtechniquestodetectunsafe online content. The system follows a modular implementationapproach,whereeachmodulesuchasuser management, comment analysis, database operations, and frontenddesignisdevelopedandtestedindependently.The backendisdevelopedusingPythonandDjango,whiletheAIbased analysis is carried out using the Gemini model integratedthroughLangChain.

3.1 User Management Implementation

TheUserManagementmoduleisimplementedusingDjango’s authentication framework. Custom models, views, and templates are designed to support user registration, login, andaccountmanagement.Theimplementationsupportsthe followingfunctionalities:

UserRegistrationwithpersonaldetails

SecureUserLoginafteradminapproval

Session-basedauthentication

PasswordresetusingOTP-basedemailverification

Userdetailssuchasname,email,mobilenumber,password, andprofileimagearestoredsecurelyinthedatabase.Django sessionsareusedtomaintainuserloginstatusandensure secureaccessthroughouttheapplication.

3.2 Comment Analysis Module Implementation

This module forms the core functionality of the Women Safety system. It allows users to submit comments or messagesforsafetyanalysis.ImplementationSteps:

Theuserentersacommentthroughthewebinterface.

The backend sends the comment to the Gemini AI model usingLangChainintegration.2025TKRCET|IT20

The AI model analyzes the text using NLP techniques to detectunsafepatterns.

The comment is classified into categories such as cyberbullying, harassment, hate speech, threats, cyberstalking,humantraffickingindicators,orsafecontent.

The classification result is returned along with a brief explanation.

Theresultisdisplayedontheuserdashboardinaclearand readableformat.

Thismoduleensuresaccurateandcontext-awaredetectionof unsafeonlinebehavior.

4. RESULTS AND PERFORMANCE ANALYSIS

This section presents the experimental results of the proposedWomenSafetyAnalyticssystem.Thesystemwas tested using multiple sample comments containing safe messagesaswellasharmfulcommentssuchascyberbullying, threats,andharassment.Theperformanceevaluationmainly focuses on correctness of classification, system response behaviour,andadmindashboardreporting.

4.1 Comment Classification Results

Thesystemsuccessfullyclassifiesuser-submittedcomments into predefined categories such as Cyberbullying & Harassment, Threats & Intimidation, Sexual Harassment, Hate Speech & Discrimination, Cyberstalking, Spam/Malicious Links, Human Trafficking Indicators, and Safe.

When a user enters a comment in the “Analyze Comment” module,theGemini2.5Flashmodelanalyzesthecontentand returnsthemostsuitablecategory.Theclassificationoutput is displayed immediately on the same page, making the systemuser-friendlyandinteractive.

4.2 User Interface and Output Display

The proposed system provides a simple and responsive interfaceforcommentsubmissionandresultvisualization. Afterclickingthe“AnalyseNow”button,theuserreceivesthe detectedcategoryalongwithaconfirmationmessage. Theoutputsectionisclearlyhighlighted,ensuringthatthe usercaneasilyunderstandwhetherthecommentissafeor harmful. Additionally, the page includes options such as AnalyzeAnotherandBacktoHome,improvingusability.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 02 | Feb 2026 www.irjet.net p-ISSN: 2395-0072

Fig. 4.2 demonstrates the successful output display after classification,confirmingthecompletionofanalysis.

4.3 Database Storage and Admin Dashboard Monitoring

All analysed comments are automatically stored in the databasealongwiththeuserinformation,detectedcategory, andtimestamp.Thisensurestransparencyandallowsfuture trackingandreporting. Theadmindashboarddisplaysboththeregisteredusersand comment analysis results in tabular format. Admins can monitor all threat detection records and manage users by activating,deactivating,ordeletingaccounts.Thisprovides complete control over system usage and improves safety monitoring.

Overall, the results confirm that the proposed system performsaccurateclassification,storesresultsefficiently,and supportsreal-timemonitoringthroughtheadmindashboard.

5. CONCLUSION

ThisprojectsuccessfullypresentsaWomenSafetyAnalytics system designed to protect women from potential threats throughreal-timemonitoringandintelligentdataanalysis.By utilizingalayeredsystemarchitecture,theproposedsolution ensuresefficientrequesthandling,securedatamanagement, and rapid alert generation during emergencies. The integrationofanalyticsenablestimelyidentificationofrisky situationsandsupportsproactivesafetymeasures.Overall, the system enhances personal security, reduces response time in critical situations, and provides a reliable technologicalsolutionthatcanbeeffectivelydeployedinrealworldenvironmentstoimprovewomen’ssafety.

6. FUTURE WORK

Infuture,thesystemcanbeextendedbyaddingsupportfor multiple languages so that harmful comments can be detectedacrossdifferentregions.Real-timemonitoringcan beimprovedbyintegratingtheapplicationwithsocialmedia platformsandmobilenotifications.AdvancedAImodelscan be incorporated to increase accuracy and detect more complex forms of abusive behavior such as sarcasm or hiddenthreats.Theprojectcanalsobeexpandedtoinclude location-based emergency alerts and direct contact with nearbyhelpcenters.Additionally,amobileapplicationcan be developed to make the system more accessible and convenientforusers.

ACKNOWLEDGEMENT

herearemanypeoplewhohelpedusdirectlyorindirectlyto completeourprojectsuccessfully.Wewouldliketotakethis opportunitytothankoneandall.Weareextremelythankful and indebted to our supervisor, Mrs. M.ARUNA Assistant Professor, Department of Information Technology, TKR College of Engineering and Technology, for his constant guidance,encouragementandmoralsupportthroughoutthe project. We are extremely thankful to Dr. R. MURUGANANTHAM, Head of the Department(I/c), Department of Information Technology, TKR College of Engineering and Technology, for the encouragement and supportthroughouttheproject.Wearesincerethankfuland gratitudetoDr.D.V.RAVISHANKAR,Principal,TKRCollege of Engineering and Technology, for all the timely support andvaluablesuggestionsduringtheperiodofourproject. Finally,wewouldalsoliketothankallthefacultyandstaffof InformationTechnologyDepartmentwhohelpedusdirectly or indirectly, parents and friends for their cooperation in completingtheprojectwork.

REFERENCES

[1] S.Sharma,A.Dubey,andR.Singh,“Smartwomensafety systemusingIoTandmachinelearning,”International JournalofAdvancedComputerScienceandApplications, vol. 10, no. 6, pp. 321–327, 2019. DOI:10.14569/IJACSA.2019.0100644

[2] M. R. Hasan, M. M. Islam, and M. A. Rahman, “An intelligentsafetysystemforwomenusingGPSandGSM technologies,” ProcediaComputerScience,vol.89, pp. 774–781, 2016.

DOI:10.1016/j.procs.2016.06.059

[3] S.PawarandA.Shinde,“WomensafetydeviceusingIoT andcloudcomputing,”IEEEInternationalConferenceon InventiveComputationTechnologies(ICICT),2018,pp. 1–5.

DOI:10.1109/ICICT43934.2018.9034317

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 02 | Feb 2026 www.irjet.net p-ISSN: 2395-0072

[4] R.Mohan,P.Karthik,andS.Ramesh,“Crimeprediction andanalysisusingmachinelearning,”IEEEInternational ConferenceonDataScienceandAnalytics,2017,pp.1–6.

DOI:10.1109/DSA.2017.8297628

[5] A.K.JainandB.Gupta,“Real-timesurveillanceandalert system for women safety using artificial intelligence,” Journal of Intelligent Systems, vol. 30, no. 1, pp. 105–118,2021.

DOI:10.1515/jisys-2019-0156

© 2026, IRJET | Impact Factor value: 8.315 | ISO 9001:2008 Certified Journal | Page 439