Scholarlink

In order to perform real-time, secure detection of anomalies in This work introduces F-HAD, a new framework that combines federal learning (FL), Full Homomorphic Security (FHE), and other technologies to allow for real-time, secure detection of problems in financial systems. zk-SNARKs. F-HAD gets 98.9% accuracy on SWIFT datasets and has an inference latency of 17.6 ms, which is 42% faster than similar secure systems. Some of the most important new features are an adaptable differential privacy mechanism, a three-layered encoding stack, adversarial reliability, and zk-verifiable audits. We also offer an overview of resources for efficiency, comparisons of post-quantum cryptography (PQC), and a fake blockchain connection for permanent proof logging. All of these features make F-HAD very useful for real-world applications.

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

AI-Driven Federated Homomorphic Anomaly Detection for Real-Time Secure Transactions

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Astha Singh,
  • Krish Awasthi,
  • Dev Singh

摘要

In order to perform real-time, secure detection of anomalies in This work introduces F-HAD, a new framework that combines federal learning (FL), Full Homomorphic Security (FHE), and other technologies to allow for real-time, secure detection of problems in financial systems. zk-SNARKs. F-HAD gets 98.9% accuracy on SWIFT datasets and has an inference latency of 17.6 ms, which is 42% faster than similar secure systems. Some of the most important new features are an adaptable differential privacy mechanism, a three-layered encoding stack, adversarial reliability, and zk-verifiable audits. We also offer an overview of resources for efficiency, comparisons of post-quantum cryptography (PQC), and a fake blockchain connection for permanent proof logging. All of these features make F-HAD very useful for real-world applications.