[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119393-en":3,"doc-seo-119393-105":29,"detail-sidebar-cat-0-en-105":89},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":11},119393,687197207057,"Sage","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","ADAPTATION OF ADVERSARIAL MACHINE LEARNING FOR TRAINING AGENTS TO COUNTER DATA ATTACKS","Adversarial Machine Learning (AML) strengthens machine-learning models against intentional data attacks by enabling robust learning under adversarial conditions. The article studies how to adapt AML techniques for intelligent agents that can autonomously detect and neutralize attacks, including data poisoning and evasion. It reviews AML foundations, major attack vectors, and agent-training methodologies, and shows that combining adversarial training with reinforcement learning improves resilience, validated through case studies in cybersecurity, autonomous systems, and finance with high detection accuracy.","# ADAPTATION OF ADVERSARIAL MACHINE LEARNINGFOR TRAINING AGENTS TO COUNTER DATA ATTACKS\n\nN.Khajynava,Z.Mutero,A.Adam  \nBelarusian State University ofInformatics and Radioelectronics,Minsk,Belarus  \nAbstract.Adversarial Machine Learning(AML)has emerged as a critical field of study.focusing on enhancingthe robustness of machine learning models against data attacks.This article explores the adaptation of AMLtechniques to train intelligent agents capable of countering various attack types.including data poisoning andevasion.We discuss the theoretical foundations of AML.prevalent attack vectors.and methodologies for agenttraining.Our findings demonstrate that integrating adversarial training with reinforcement leanung signuificantlyimproves model resilience.ensuring the security of machine learning applications.The proposed approach isvalidated through case studies in cybersecurity,autonomous systems,and finance.Experiments show that AML-trained agents achieve up to 92%attack detection accuracy,reducing risks in autonomous systens by40%.Keywords:Adversarial Machine Learning(AML):adversarial example generation;robust model training,datapoisoning attacks:evasion resistance:AI security:reinforcement learning defense:adversarial robustness:machine learning:multi-agent systems(MAS).  \n## Introduction\n\nThe rapid integration of machine learning(ML)into critical sectors such as healthcare,finance,and autonomous systems has underscored its transformative potential.However,thisprogress is accompanied by growing vulnerabilities to adversarial attacks,where maliciousactors manipulate input data to deceive models [2].Adversarial Machine Learning(AML)addresses these threats by developing techniques to fortify models against intentional datadistortions.  \nA key challenge lies in the dynamic nature of attacks.Traditional ML systems,designedfor static environments,often fail to adapt to evolving adversarial strategies.For instance,evasion attacks,which perturb input data during inference,can mislead autonomous vehiclesinto misclassifying road signs [3].Similarly,poisoning attacks corrupt training datasets,causing models to learn biased or incorrect patterns [4].These vulnerabilities highlightthe need for adaptive defense mechanisms.  \nThis article proposes a paradigm shift:training intelligent agents using AML principlesto autonomously detect and neutralize data attacks.Unlike static models,agents can leveragereinforcement learning(RL)to dynamically adjust their strategies in response to adversarialbehavior.By integrating adversarial training -where models are exposed to perturbed inputsduring learning-agents develop inherent resistance to manipulation.This hybrid approachbridges the gap between robustness and adaptability,offering a scalable solution for securingMLapplications.  \n## Main Part\n\nAdversarial Machine Learning(AML)is a side branch of ML that has become thetheoretical basis for developing tools that can interfere with the operation of ML-basedsystems.The term Adversarial Machine Learning is still rarely found in Russian-languagetexts,it is translated as \"cOCT93aTeJIbHOe MaIHHHoe o6yueHue\",but more accurately,theword adversarial has meanings from the series antagonistic,confrontational,or opposing,  \nso by analogy with malware,it can be translated as《BpeIoHOCHoe MaIHHHoe o6yueHHe》.The discovery of the theoretical possibility of the existence of AML and the first publicationson this topic date back to 2004.The history of AML and an analysis of the current stateof affairs can be found in the article“Wild Patterns:Ten Years After the Rise of AdversarialMachine Learning”by two Italian researchers Battista Biggioa and Fabio Rolia,publishedin 2018[1].  \nAdversarial Machine Learning(AML)is rooted in the interplay between attack anddefense strategies.At its core,AML studies how models can be deceived by carefully craftedinputs,known as adversarial examples,and how to mitigate such threats [2].Gradient-basedmethods,such as the Fast Gradient Sign ","cbCaitFB221Tljzy","https://ap.wps.com/l/cbCaitFB221Tljzy","pdf",161933,1,3,"English","en",105,"# Introduction\n# Main Part\n## Core concepts of AML\n## Gradient-based generation of adversarial examples\n## Agent-based adaptation with reinforcement learning\n## Applications in cybersecurity\n## Applications in autonomous systems\n## Applications in financial systems\n## Open challenges and trade-offs","[{\"question\":\"What problem does adversarial machine learning address for agent-based systems?\",\"answer\":\"It targets vulnerabilities where malicious actors manipulate inputs or training data to deceive ML models. For agents, the goal is to counter evolving attack strategies by adapting behavior during training and operation.\"},{\"question\":\"How do adversarial examples like FGSM and PGD relate to AML defenses?\",\"answer\":\"Gradient-based methods such as FGSM and PGD generate adversarial examples by using model gradients to create small, human-imperceptible perturbations that can mislead models. AML then studies how to mitigate these threats through robust training and defenses.\"},{\"question\":\"Why combine adversarial training with reinforcement learning when training agents?\",\"answer\":\"Adversarial training exposes models to perturbed inputs so they learn inherent resistance. Reinforcement learning allows agents to dynamically adjust strategies in response to adversarial behavior, improving robustness and adaptability.\"}]","ADAPTATION OF ADVERSARIAL MACHINE LEARNING FOR TRAINING AGENTS TO COUNTER DATA ATTACKS | PDF",1785724075,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":84,"head_meta":86,"extra_data":88,"updated_unix":28},"adaptation-of-adversarial-machine-learning-for-training-agents-to-counter-data-attacks","",{"@graph":35,"@context":83},[36,52,66],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,49],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":21},"https://docshare.wps.com/document/research-report/",{"item":50,"name":13,"@type":42,"position":51},"https://docshare.wps.com/document/adaptation-of-adversarial-machine-learning-for-training-agents-to-counter-data-attacks/119393/",4,{"url":50,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":23,"description":14,"dateModified":60,"datePublished":60,"encodingFormat":59,"isAccessibleForFree":61,"interactionStatistic":62},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":40,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":63,"interactionType":64,"userInteractionCount":4},"InteractionCounter",{"@type":65},"ViewAction",{"@type":67,"mainEntity":68},"FAQPage",[69,75,79],{"name":70,"@type":71,"acceptedAnswer":72},"What problem does adversarial machine learning address for agent-based systems?","Question",{"text":73,"@type":74},"It targets vulnerabilities where malicious actors manipulate inputs or training data to deceive ML models. For agents, the goal is to counter evolving attack strategies by adapting behavior during training and operation.","Answer",{"name":76,"@type":71,"acceptedAnswer":77},"How do adversarial examples like FGSM and PGD relate to AML defenses?",{"text":78,"@type":74},"Gradient-based methods such as FGSM and PGD generate adversarial examples by using model gradients to create small, human-imperceptible perturbations that can mislead models. AML then studies how to mitigate these threats through robust training and defenses.",{"name":80,"@type":71,"acceptedAnswer":81},"Why combine adversarial training with reinforcement learning when training agents?",{"text":82,"@type":74},"Adversarial training exposes models to perturbed inputs so they learn inherent resistance. Reinforcement learning allows agents to dynamically adjust strategies in response to adversarial behavior, improving robustness and adaptability.","https://schema.org",{"og:url":50,"og:type":85,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":87,"canonical":50},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":90},[91,95,99,103,108,113,118,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":92,"show_sort_weight":93,"slug":94},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":96,"show_sort_weight":97,"slug":98},"Literature",80,"literature",{"id":51,"doc_module":4,"doc_module_name":45,"category_name":100,"show_sort_weight":101,"slug":102},"Exam",70,"exam",{"id":104,"doc_module":4,"doc_module_name":45,"category_name":105,"show_sort_weight":106,"slug":107},5,"Comic",60,"comic",{"id":109,"doc_module":4,"doc_module_name":45,"category_name":110,"show_sort_weight":111,"slug":112},6,"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":104,"slug":136},19,"General","general"]