Scholarlink

<p>Artificial intelligence (AI) has revolutionized diagnostic accuracy and patient care by enhancing the accuracy and efficiency of medical imaging. AI applications such as deep neural networks are able to process complex medical images with remarkable accuracy to provide clinicians with valuable insight into an individual’s disease state and help in early disease detection. However, the increasing use of AI in medical imaging has raised many challenges in terms of protection of data especially protection of patients’ confidential information. The majority of existing encryption techniques require decryption of all information before processing with AI. This method can expose sensitive information in an AI image to potential risks of loss or theft. Moreover, modern cryptographic algorithms such as RSA or Elliptic Curve Cryptography (ECC) are susceptible to quantum attacks, highlighting the need for a novel cryptographic architecture capable of ensuring secure data encryption in the quantum era. To address the above challenges, we propose a lattice-based zero-knowledge STARK (Scalable Transparent Argument of Knowledge) framework that is highly adaptable and designed to ensure secure AI-aided medical diagnostics. The framework uses the hardness of lattice problems such as Short Integer Solution (SIS) and Module Learning with Errors (MLWE) to form quantum-resilient commitments while enabling the verification of a diagnostic outcome by Zero Knowledge Proof (ZKP). By using Ring-LWE-based hash functions with Merkle tree commitment structures, we achieve the authentication of possible artificial intelligence outputs in the end without exposing medically sensitive images. Experimental evaluation shows that the proposed solution outperforms traditional encryption-based methods by offering enhanced security, stronger privacy guarantees and improved compliance with healthcare data protection standards. This framework allows to build trust and privacy-preserving AI ecosystem in healthcare, thereby strengthening the integrity and reliability of healthcare diagnostics.</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

Revolutionizing medical imaging privacy with lattice-based zk-STARKs in AI driven healthcare

  • 期刊论文

  • 发表日期：

    2026年2月23日

查看全文
  • Trisha Bhavini,
  • Senthil Prakash P.N,
  • Ravulapati Shipreethi,
  • Kamkanala Keerthi

摘要

Artificial intelligence (AI) has revolutionized diagnostic accuracy and patient care by enhancing the accuracy and efficiency of medical imaging. AI applications such as deep neural networks are able to process complex medical images with remarkable accuracy to provide clinicians with valuable insight into an individual’s disease state and help in early disease detection. However, the increasing use of AI in medical imaging has raised many challenges in terms of protection of data especially protection of patients’ confidential information. The majority of existing encryption techniques require decryption of all information before processing with AI. This method can expose sensitive information in an AI image to potential risks of loss or theft. Moreover, modern cryptographic algorithms such as RSA or Elliptic Curve Cryptography (ECC) are susceptible to quantum attacks, highlighting the need for a novel cryptographic architecture capable of ensuring secure data encryption in the quantum era. To address the above challenges, we propose a lattice-based zero-knowledge STARK (Scalable Transparent Argument of Knowledge) framework that is highly adaptable and designed to ensure secure AI-aided medical diagnostics. The framework uses the hardness of lattice problems such as Short Integer Solution (SIS) and Module Learning with Errors (MLWE) to form quantum-resilient commitments while enabling the verification of a diagnostic outcome by Zero Knowledge Proof (ZKP). By using Ring-LWE-based hash functions with Merkle tree commitment structures, we achieve the authentication of possible artificial intelligence outputs in the end without exposing medically sensitive images. Experimental evaluation shows that the proposed solution outperforms traditional encryption-based methods by offering enhanced security, stronger privacy guarantees and improved compliance with healthcare data protection standards. This framework allows to build trust and privacy-preserving AI ecosystem in healthcare, thereby strengthening the integrity and reliability of healthcare diagnostics.