TL;DR
This paper introduces a novel reversible data hiding (RDH) scheme for encrypted medical images, achieving high embedding capacity and visual quality. To address integrity and traceability concerns, they integrate this RDH scheme with a blockchain system, creating a secure and auditable platform for medical image exchange.
What Problem Does This Paper Address?
The increasing digitization of healthcare means medical images are frequently shared, making them vulnerable to malicious attacks. While encryption protects image content, it doesn't inherently allow for embedding private patient data within the image itself. Reversible data hiding on encrypted images (RDHEI) offers a solution by embedding private information (like medical records) into encrypted medical images, allowing for both secure communication and diagnostic support. However, existing RDHEI methods often struggle with low embedding capacity and lack robust mechanisms for ensuring data integrity and traceability. This paper aims to overcome these limitations by proposing a more efficient RDHEI technique and integrating it with a blockchain for enhanced security and audibility.
Methods
The proposed solution has two main components:
- RDHEI Scheme:
- Block-wise Encryption: The cover image is first divided into 3x3 blocks. These blocks are then permuted using a key (Key-I) to scramble spatial relationships. Subsequently, a stream cipher (using Key-II) is applied to encrypt each block. This approach aims to improve security and prepare the image for data embedding.
- Histogram Shifting: After encryption, a preprocessing step (using an Exclusive-OR operation) creates "vacating bands" in the histogram. A histogram shifting technique is then employed to embed secret data into these bands. The threshold for this shifting (gth) is adjustable, allowing for a trade-off between embedding capacity and visual quality.
- Reversibility: The process is designed to be fully reversible. The original image can be perfectly reconstructed after extracting the secret data, and the secret data can be extracted without altering the original image.
- Blockchain Integration:
- Consortium Blockchain: A consortium blockchain is proposed, assuming participants (like hospitals and research institutes) are trusted. This allows for a lightweight consensus mechanism (like PBZT) instead of computationally expensive ones like Proof-of-Work.
- Integrity and Traceability: The hash value of the encrypted medical image (Ciphered Steganography Medical Image - CSMI) is generated and stored on the blockchain, along with the hash of the previous block. This creates an immutable ledger. Any attempt to tamper with the CSMI will result in a different hash value, which can be easily detected by comparing it with the stored hash on the blockchain. This ensures the integrity of the medical images and provides a traceable history of their communication.
Datasets
The paper uses two sets of datasets for evaluation:
- Standard Test Images: Six common grayscale images (Airplane, Baboon, Boat, Lena, Peppers, Sailboat) of size 512x512 pixels.
- Medical Images: Two MRI medical images of size 256x256 pixels, obtained from MIDAS/National Alliance for Medical Image Computing (NAMIC).
Key Insights and Ideas
- Synergistic Security: Combining RDHEI with blockchain offers a multi-layered security approach. RDHEI secures the image content and allows private data embedding, while blockchain provides an auditable trail and integrity verification.
- Flexible Trade-offs: The adjustable threshold (gth) in the histogram shifting mechanism allows users to dynamically balance embedding capacity and visual quality, which is crucial for different diagnostic needs.
- Stream Cipher Advantage: Using a stream cipher for encryption, in conjunction with block permutation, is presented as an improvement over previous methods, potentially offering better security and preparing the image more effectively for data hiding.
- Blockchain for Healthcare Auditing: The application of a consortium blockchain in healthcare for medical image integrity and traceability is a significant step towards building trust in digital health systems.
Key Results
- High Embedding Rate: The proposed RDHEI scheme achieves an embedding rate of over 0.8 bits per pixel (bpp) for typical medical images, and up to 0.85 bpp for medical images at the highest threshold. This is noted as being higher than some existing schemes.
- Good Visual Quality: The reconstructed images, even with fast recovery (where some information might be discarded), exhibit good visual quality, as indicated by high PSNR and SSIM values (e.g., PSNR > 30 dB and SSIM > 0.98 for most cases).
- Efficient Encryption: The use of simple Exclusive-OR operations in the encryption and preprocessing steps makes the scheme computationally efficient.
- Robustness: The image encryption scheme is claimed to be robust against chosen-plaintext attacks due to the corruption of block correspondence and the use of distinct keys for block encryption.
Limitations and Future Work
- Trusted Participants: The blockchain system relies on a consortium model, assuming participants are trusted. The security of this model depends on the trustworthiness of these pre-defined members.
- Key Management: While keys are mentioned, the paper doesn't delve deeply into the practical aspects of secure key management for multiple users and devices, which is a common challenge in such systems.
- Performance with Complex Images: The paper notes that the PSNR value is dependent on image features, with the "Baboon" image showing the lowest PSNR, suggesting potential challenges with highly complex textures.
- Future Work: The authors suggest that the proposed RDHEI blockchain system can be implemented with popular blockchain frameworks like Hyperledger Fabric and Ethereum, and that industry-grade block ciphers like AES can be used for initial encryption. Further research could explore more advanced consensus mechanisms or decentralized approaches for broader applicability.
Summary updated