Scholarlink

<p>With the rapid digital transformation in the banking industry, it has become imperative to transform cyber defense strategies into intelligent and adaptive ones. This paper presents a conceptual study of adversarial machine learning (AML) that would give rise to a system which we shall call Automated Cyber Threat Protection (ACTP). It is proactive in its approach toward detecting and mitigating threats within a banking environment. Reinforcement learning on ethical hacking simulation infused with adversarial training are being used under ACTP to dynamically model known as well as constantly evolving threat vectors. Unlike static systems, ACTP runs in a closed-loop design by continuously injecting synthetic attacks for further improvement of detection models in real time. The prototype was validated with both synthetic and real-world banking cyberattack datasets and delivered an accuracy of 88.2% with a false positive rate of only 5.1%. The prototype reduced false negatives by 42%. Further average threat response latency is improved by 61% over baseline intrusion detection systems. This work shall have two main contributions: (i) the development of a unified, AI-based coherent threat detection framework under conditions of adversarial attacks; and (ii) empirical validation in practice that under such conditions this framework can be implemented to increase cybersecurity resilience across financial infrastructures. By bringing together predictive intelligence with automated prevention, ACTP is a stride toward implementing autonomous cybersecurity for the financial sector.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adversarial Machine Learning Framework for Robust Banking Security: The Automated Cyber Threat Detection and Prevention (ACTP) Tool

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Harish Padmanaban,
  • Yogesh Kumar Sharma,
  • Purushottam Sharma,
  • Chaman Verma

摘要

With the rapid digital transformation in the banking industry, it has become imperative to transform cyber defense strategies into intelligent and adaptive ones. This paper presents a conceptual study of adversarial machine learning (AML) that would give rise to a system which we shall call Automated Cyber Threat Protection (ACTP). It is proactive in its approach toward detecting and mitigating threats within a banking environment. Reinforcement learning on ethical hacking simulation infused with adversarial training are being used under ACTP to dynamically model known as well as constantly evolving threat vectors. Unlike static systems, ACTP runs in a closed-loop design by continuously injecting synthetic attacks for further improvement of detection models in real time. The prototype was validated with both synthetic and real-world banking cyberattack datasets and delivered an accuracy of 88.2% with a false positive rate of only 5.1%. The prototype reduced false negatives by 42%. Further average threat response latency is improved by 61% over baseline intrusion detection systems. This work shall have two main contributions: (i) the development of a unified, AI-based coherent threat detection framework under conditions of adversarial attacks; and (ii) empirical validation in practice that under such conditions this framework can be implemented to increase cybersecurity resilience across financial infrastructures. By bringing together predictive intelligence with automated prevention, ACTP is a stride toward implementing autonomous cybersecurity for the financial sector.