Privacy-preserving machine learning using EtC images

Kawamura, Ayana; Kinoshita, Yuma; Kiya, Hitoshi

doi:10.1117/12.2566272

Cited by 6 publications

(4 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Kawamura et al (2019) proposed a privacy-conserving ML system with encryption-then-compression (EtC) images. A new property enables us to adapt EtC images without any degradation in classification quality to some ML algorithms.…”

Section: Related Workmentioning

confidence: 99%

Emphasizing privacy and security of edge intelligence with machine learning for healthcare

Rajendran

Mathivanan

Jayagopal

et al. 2021

IJICC

View full text Add to dashboard Cite

PurposeArtificial Intelligence (AI) has surpassed expectations in opening up different possibilities for machines from different walks of life. Cloud service providers are pushing. Edge computing reduces latency, improving availability and saving bandwidth.Design/methodology/approachThe exponential growth in tensor processing unit (TPU) and graphics processing unit (GPU) combined with different types of sensors has enabled the pairing of medical technology with deep learning in providing the best patient care. A significant role of pushing and pulling data from the cloud, big data comes into play as velocity, veracity and volume of data with IoT assisting doctors in predicting the abnormalities and providing customized treatment based on the patient electronic health record (EHR).FindingsThe primary focus of edge computing is decentralizing and bringing intelligent IoT devices to provide real-time computing at the point of presence (PoP). The impact of the PoP in healthcare gains importance as wearable devices and mobile apps are entrusted with real-time monitoring and diagnosis of patients. The impact edge computing of the PoP in healthcare gains significance as wearable devices and mobile apps are entrusted with real-time monitoring and diagnosis of patients.Originality/valueThe utility value of sensors data improves through the Laplacian mechanism of preserved PII response to each query from the ODL. The scalability is at 50% with respect to the sensitivity and preservation of the PII values in the local ODL.

show abstract

Section: Related Workmentioning

confidence: 99%

Emphasizing privacy and security of edge intelligence with machine learning for healthcare

Rajendran

Mathivanan

Jayagopal

et al. 2021

IJICC

View full text Add to dashboard Cite

show abstract

“…Ergun et al [12] proposed a Privacy Preserved Face Recognition (PPFR) framework, which encrypted the plaintext by random cryptographic keys based on the continuous chaotic system [22] and directly recognized the encrypted data. Block-based encryption methods were popular in this area, such as Combined Cat Map (CCM) [23] based on hybrid chaotic maps and dynamic random growth technique, encryption method of histograms of oriented gradients (HOG) features [24], and Encryption-then-Compression (EtC) framework [25]. Moreover, Chuman et al [26] presented a block scrambling-based encryption scheme to enhance the security of EtC framework with JPEG compression.…”

Section: A Visual Information Hidingmentioning

confidence: 99%

“…3) Comparison with Others: To evaluate the effectiveness, we compare the proposed MIAIS framework with some other visual information hiding methods on AFD datasets. Take the face recognition result on secret image as upper bound, the competing methods include Combined Cat Map (CCM) [23], Learnable Encryption (LE) [27], PBIE [29], Encryption-Then-Compression (EtC) [25], Extended Learnable Encryption (ELE) [13], EM-HOG [24], Defeating Image Obfuscation (DIO) [30], Privacy Preserving Face Recognition (PPFR) [12] and a naive Block-Wise pixel Shuffle (BWS). BWS only naively shuffles the 4 × 4 pixel blocks without consideration for privacy protection.…”

Section: ) Ablation Studymentioning

confidence: 99%

Multitask Identity-Aware Image Steganography via Minimax Optimization

Cui¹,

Zhang²,

Li³

et al. 2021

Preprint

View full text Add to dashboard Cite

High-capacity image steganography, aimed at concealing a secret image in a cover image, is a technique to preserve sensitive data, e.g., faces and fingerprints. Previous methods focus on the security during transmission and subsequently run a risk of privacy leakage after the restoration of secret images at the receiving end. To address this issue, we propose a framework, called Multitask Identity-Aware Image Steganography (MIAIS), to achieve direct recognition on container images without restoring secret images. The key issue of the direct recognition is to preserve identity information of secret images into container images and make container images look similar to cover images at the same time. Thus, we introduce a simple content loss to preserve the identity information, and design a minimax optimization to deal with the contradictory aspects. We demonstrate that the robustness results can be transferred across different cover datasets. In order to be flexible for the secret image restoration in some cases, we incorporate an optional restoration network into our method, providing a multitask framework. The experiments under the multitask scenario show the effectiveness of our framework compared with other visual information hiding methods and state-of-the-art high-capacity image steganography methods.

show abstract

“…Therefore, encrypted images can be directly applied to image processing algorithms. Some encryption methods have been studied for applying encrypted images to traditional machine learning algorithms, such as support vector machine (SVM), under the use of the kernel trick [21], [22], but they cannot be applied to DNNs [15]. There are four perceptual encryption methods [10], [14]- [16] for privacy-preserving DNNs.…”

Section: Related Work a Visual Information Protectionmentioning

confidence: 99%

Visual Security Evaluation of Learnable Image Encryption Methods against Ciphertext-only Attacks

Sirichotedumrong¹,

Kiya²

2020

Preprint

Self Cite

View full text Add to dashboard Cite

Various methods for protecting visual information have been proposed for privacy-preserving deep neural networks (DNNs). In contrast, methods of attack against such protection methods have been simultaneously developed. In this paper, we evaluate state-of-the-art visual-protection methods for privacypreserving DNNs in terms of visual security against ciphertextonly attacks (COAs). We focus on the brute-force attack, feature reconstruction attack (FR-Attack), inverse transformation attack (ITN-Attack), and GAN-based attack (GAN-Attack), which have been proposed to reconstruct the visual information of plain images from visually protected images. The details of the various attacks are first summarized, and the visual security of the protection methods is then evaluated. Experimental results demonstrate that most of the protection methods, including pixelwise encryption, are not robust enough against GAN-Attack, while a few are robust enough against it.

show abstract

Privacy-preserving machine learning using EtC images

Cited by 6 publications

References 27 publications

Emphasizing privacy and security of edge intelligence with machine learning for healthcare

Emphasizing privacy and security of edge intelligence with machine learning for healthcare

Multitask Identity-Aware Image Steganography via Minimax Optimization

Visual Security Evaluation of Learnable Image Encryption Methods against Ciphertext-only Attacks

Contact Info

Product

Resources

About