Maintaining Data Privacy Research Articles

A Secure and Lightweight Multi-Party Private Intersection-Sum Scheme over a Symmetric Cryptosystem

A private intersection-sum (PIS) scheme considers the private computing problem of how parties jointly compute the sum of associated values in the set intersection. In scenarios such as electronic voting, corporate credit investigation, and ad conversions, private data are held by different parties. However, despite two-party PIS being well-developed in many previous works, its extended version, multi-party PIS, has rarely been discussed thus far. This is because, depending on the existing works, directly initiating multiple two-party PIS instances is considered to be a straightforward way to achieve multi-party PIS; however, by using this approach, the intersection-sum results of the two parties and the data only belonging to the two-party intersection will be leaked. Therefore, achieving secure multi-party PIS is still a challenge. In this paper, we propose a secure and lightweight multi-party private intersection-sum scheme called SLMP-PIS. We maintain data privacy based on zero sharing and oblivious pseudorandom functions to compute the multi-party intersection and consider the privacy of associated values using arithmetic sharing and symmetric encryption. The security analysis results show that our protocol is proven to be secure in the standard semi-honest security model. In addition, the experiment results demonstrate that our scheme is efficient and feasible in practice. Specifically, when the number of participants is five, the efficiency can be increased by 22.98%.

Hybrid cloud storage system with enhanced multilayer cryptosystem for secure deduplication in cloud

Data deduplication is a crucial technique in the field of data compression that aims to eliminate redundant copies of recurring data. This technique has gained significant popularity in the realm of cloud storage due to its ability to effectively reduce storage requirements and optimize bandwidth utilization. To ensure the safeguarding of sensitive data while simultaneously facilitating deduplication, researchers have put forth the concept of convergent encryption as a potential solution. This technique involves encrypting the data prior to its outsourcing, thereby enhancing the confidentiality of the information. In this work, an earnest endeavor is undertaken to formally tackle the issue of authorized data deduplication, with the aim of enhancing data security. Our approach combines the Diffie-Hellman algorithm and symmetrical external decision to protect and popularize information, ensuring end-to-end encryption to encourage user adoption of cloud storage. The proposed model employs block-level deduplication and guarantees the randomness of ciphertexts by generating encryption keys using the Diffie-Hellman algorithm. This method effectively counters both internal and external brute-force attacks, enhancing data security while reducing computational costs. An extensive experimentation is carried out to demonstrate that our approach is particularly beneficial in scenarios with multiple privilege sets. Overall, the proposed model offers an elaborate framework that maintains data privacy and strengthens security measures, contributing to a more efficient and secure cloud-based document search.

Translating Usage Control Policies to Semantic Rules: A Model using OrBAC and SWRL

The increasing volume of data in various environments such as IoT and the need to maintain data privacy and security have led to the development of usage control models. Usage control policies are models that enable fine-grained access control over data by enforcing restrictions on how users can use the data. Semantic mechanisms, on the other hand, use context and meaning to identify potential security threats and prevent them from accessing sensitive information. Although not widely explored, merging these two techniques could create an efficient mechanism to help ensure the confidentiality, integrity, and availability of critical data and resources. This paper aims to encourage this research path by proposing a translation model that converts usage control rules into SWRL. In particular, we consider during our approach the notions of contexţ permission and prohibition. The proposition is validated by constructing a multi-layer proof of concept that use ontologies and OWL for implementing the translation model. Furthermore, to ascertain the practicality of our approach, a time processing evaluation is conducted, and the results are found to be satisfactory.

SENSIBLE: SEquestered aNd SynergIstic BLockchain Ecosystem

AbstractHealth care interoperability unfolds the way for personalized health care services at a reduced cost. Furthermore, a decentralized system holds the promise to prevent compromises such as cyber‐attacks due to data breaches. Hence, there is a need for a framework that seamlessly integrates and shares data across the system stakeholders. We propose SEquestered aNd SynergIstic BLockchain Ecosystem (SENSIBLE), a blockchain‐powered, knowledge‐driven data‐sharing framework that gives patients complete control of their medical history and can extract rich information hidden in it using knowledge graphs (KGs). By incorporating both blockchain and KGs, we can provide a platform for secure data sharing among stakeholders by maintaining data privacy and integrity through data authentication and robust data integration. We present a Proof‐of‐Concept of the SENSIBLE network with Ethereum to share dynamic knowledge across stakeholders. Dynamic knowledge generation on the blockchain provides a two‐fold advantage of cooperation and communication amongst the stakeholders in the health care ecosystem. This leads to operational ease through sharing relevant portions of complex information while also ensuring the isolation of sensitive medical data.

Blockchain-Based Reliable and Efficient Certificateless Signature for IIoT Devices

Nowadays, the Industrial Internet of Things (IIoT) has remarkably transformed our personal lifestyles and society operations into a novel digital mode, which brings tremendous associations with all walks of life, such as intelligent logistics, smart grid, and smart city. Moreover, with the rapid increase of IIoT devices, a large amount of data is swapped between heterogeneous sensors and devices every moment. This trend increases the risk of eavesdropping and hijacking attacks in communication channels, so maintaining data privacy and security becomes two notable concerns at present. Recently, based on the mechanism of the Schnorr signature, a more secure and lightweight certificateless signature (CLS) protocol is popular for the resource-constrained IIoT protocol design. Nevertheless, we found most of the existing CLS schemes are susceptible to several common security weaknesses such as man-in-the-middle attacks, key generation center compromised attacks, and distributed denial of service attacks. To tackle the challenges mentioned previously, in this article, we propose a novel pairing-free certificateless scheme that utilizes the state-of-the-art blockchain technique and smart contract to construct a novel reliable and efficient CLS scheme. Then, we simulate the Type-I and Type-II adversaries to verify the trustworthiness of our scheme. Security analysis as well as performance evaluation outcomes prove that our design can hold more reliable security assurance with less computation cost (i.e., reduced by around 40.0% at most) and communication cost (i.e., reduced by around 94.7% at most) than other related schemes.

Brakerski‐Gentry‐Vaikuntanathan fully homomorphic encryption cryptography for privacy preserved data access in cloud assisted Internet of Things services using glow‐worm swarm optimization

AbstractHealth monitoring sensors are widely available to monitor patients' health remotely. The collected sensor data are sent to cloud storage and processed in a remote health monitoring system. The Internet of Things (IoT) with cloud support offers a promising answer to data exploding concerns for the ability to constrain specific gadgets. However, IoT faces many security issues when sharing data between two users because of the cloud's leverage nature. In the case of a private cloud, simple encryption techniques can ensure security. But, in the case of a public cloud, maintaining data privacy is a significant issue. To overcome this concern, an optimized Brakerski‐Gentry‐Vaikuntanathan fully homomorphic encryption (BGV‐FHE) encryption method for secured data mutuality is proposed in this article. In this method, IoT medical data are initially acquired and then encrypted using two key schemes, and the data are stored in cloud storage. The main objective of these key schemes is to select optimal key parameters. Key parameters are chosen using the glow‐worm swarm optimization and used for the encryption phase. Storing data in the cloud using a public key provides access to requested users. Meanwhile, sensitive data are encrypted using a private key and cannot be accessed unless authenticated. This proposed approach has two levels of encryption, and thus it is an efficient data secured protocol for cloud resources. Implementing the optimized BGV‐FHE scheme based encryption scheme in Python, two datasets result in an accuracy level of 77% and 91%, respectively.

Blockchain Technology in Healthcare

Blockchain technology has recently been at the center stage of academic discussions. This is owing to its use in the financial sector and the rise of cryptocurrencies like Bitcoin. In essence, digital ledger technology (DLT) is shared digital data spread across multiple sites. A blockchain is a type of DLT consisting of a growing list of records, called blocks, securely linked with a time stamp. Maintaining data privacy and security is one of the core benefits of using the blockchain network. In healthcare, efficient data sharing is desirable with secure long-term access to data. In recent years, numerous applications built on blockchain infrastructure have been introduced in the healthcare sector. In this narrative review, we intend to introduce the science of blockchain technology, its core concepts, and its application to healthcare. We also describe current instances of its application in the healthcare sector, its application from the Indian perspective, and its limitations and future possibilities.

Multi-institutional distributed data networks for real-world evidence about medical devices: building unique device identifiers into longitudinal data (BUILD).

ObjectivesTo support development of a robust postmarket device evaluation system using real-world data (RWD) from electronic health records (EHRs) and other sources, employing unique device identifiers (UDIs) to link to device information.MethodsTo create consistent device-related EHR RWD across 3 institutions, we established a distributed data network and created UDI-enriched research databases (UDIRs) employing a common data model comprised of 24 tables and 472 fields. To test the system, patients receiving coronary stents between 2010 and 2019 were loaded into each institution’s UDIR to support distributed queries without sharing identifiable patient information. The ability of the system to execute queries was tested with 3 quality assurance checks. To demonstrate face validity of the data, a retrospective survival study of patients receiving zotarolimus or everolimus stents from 2012 to 2017 was performed using distributed analysis. Propensity score matching was used to compare risk of 6 cardiovascular outcomes within 12 months postimplantation.ResultsThe test queries established network functionality. In the analysis, we identified 9141 patients (Mercy = 4905, Geisinger = 4109, Intermountain = 127); mean age 65 ± 12 years, 69% males, 23% zotarolimus. Separate matched analyses at the 3 institutions showed hazard ratio estimates (zotarolimus vs everolimus) of 0.85–1.59 for subsequent percutaneous coronary intervention (P = .14–.52), 1.06–2.03 for death (P = .16–.78) and 0.94–1.40 for the composite endpoint (P = .16–.62).DiscussionThe analysis results are consistent with clinical studies comparing these devices.ConclusionThis project shows that multi-institutional data networks can provide clinically relevant real-world evidence via distributed analysis while maintaining data privacy.

Chaotic-Based Encryption Algorithm using Henon and Logistic Maps for Fingerprint Template Protection

Fingerprint is a reliable user authentication method as it is unique to individual users that makes it efficient for authenticating users. In a fingerprint authentication system, user fingerprint information is stored in databases in an image format known as a fingerprint template. Although fingerprint is reliable, the templates stored in the database are exposed to security threats either during the data transmission process over the network or in storage. Therefore, there is a need to protect the fingerprint template, especially in unsecured networks to maintain data privacy and confidentiality. Many past studies proposed fingerprint template protection (FTP) using chaotic-based encryption algorithms that are more suitable to secure images than conventional encryption such as DES, AES, and RSA. The chaotic-based encryption algorithms have been improved a lot in terms of their robustness. However, the robustness of the algorithm caused a trade-off to encryption speed where it remains an issue in FTP. Hence, this study aims to improve the limitations found in the existing chaotic-based encryption algorithms for FTP by improving its encryption speed using Henon and Logistic map. A series of simulations were conducted using MATLAB to evaluate the performance of the proposed chaotic-based encryption algorithm for FTP through different analyses covering key sensitivity, histogram, correlations, differential, information entropy, and encryption/decryption speed. The performance proposed encryption algorithm was promising which could be a starting point for detailed analysis and implementation in real application domains.

Novel technical and privacy-preserving technology for artificial intelligence in ophthalmology.

The application of artificial intelligence (AI) in medicine and ophthalmology has experienced exponential breakthroughs in recent years in diagnosis, prognosis, and aiding clinical decision-making. The use of digital data has also heralded the need for privacy-preserving technology to protect patient confidentiality and to guard against threats such as adversarial attacks. Hence, this review aims to outline novel AI-based systems for ophthalmology use, privacy-preserving measures, potential challenges, and future directions of each. Several key AI algorithms used to improve disease detection and outcomes include: Data-driven, imagedriven, natural language processing (NLP)-driven, genomics-driven, and multimodality algorithms. However, deep learning systems are susceptible to adversarial attacks, and use of data for training models is associated with privacy concerns. Several data protection methods address these concerns in the form of blockchain technology, federated learning, and generative adversarial networks. AI-applications have vast potential to meet many eyecare needs, consequently reducing burden on scarce healthcare resources. A pertinent challenge would be to maintain data privacy and confidentiality while supporting AI endeavors, where data protection methods would need to rapidly evolve with AI technology needs. Ultimately, for AI to succeed in medicine and ophthalmology, a balance would need to be found between innovation and privacy.

Electronic Health Record Monitoring System and Data Security Using Blockchain Technology

Bangladesh should have owned a decentralized medical record server. We face a lot of issues, such as doctor’s appointments, report organization in one spot, and report follow-ups. People now bring a large number of papers to the doctor’s chamber. They carry prescriptions, reports, and X-ray files, among other things. It complicates everyone’s life as a result. All of the reports must be reviewed by doctors on a regular basis. It is difficult to read old reports on a regular basis, and patients do not receive the correct medications or treatment. Doctors also find it extremely difficult to comprehend handwritten prescriptions. Data security, authenticity, time management, and other areas of data administration are dramatically improved when blockchain (smart contract) technology is linked with standard database management solutions. Blockchain is a groundbreaking, decentralized technology that protects data from unauthorized access. After smart contracts are implemented, the management will be satisfied with the patients. As a result, maintaining data privacy and accountability in the system is tough. It signifies that the information is only accessible to those who have been authenticated. This study focuses on limiting third-party engagement in medical health data and improving data security. Throughout the process, this will improve accessibility and time efficiency. People will feel safer during the payment procedure, which is the most significant benefit. A smart contract and a peer-to-peer encrypted technology were used. The hacker will not be able to gain access to this system since this document uses an immutable ledger. They will not be able to change any of the data if they gain access to the system. If the items are found to be defective, the transaction will be halted. Transaction security will be a viable option for recasting these problems using cryptographic methodologies. We developed a website where patients and doctors will both benefit because of the use of blockchain technology to ensure the security of medical data. We have different profiles for doctors and patients. In the patient profile, they can create their own account by using a unique address, name, and age. This unique address will be created from the genesis block. The unique address is completely private to the owner, who will remain fully secure in our network. After creating an account, the patient can view the doctors’ list and they can upload their medical reports such as prescriptions and X-rays. All the records uploaded by the patient will be stored on our local server (Ganache). The records are stored as hashed strings of the data. Those files will also have a unique address, and it will be shown in the patient profile. After granting access, the doctors will be able to view their records in the respective doctor’s profile. For accessing the options such as uploading, viewing, or editing the data, Ethereum currency (a fee) will have to be paid in order to complete the request. On the other hand, doctors can enter their profile using their name and unique address. After logging in, they can view their name, unique address, and the list of patients that have granted access to the doctor to view their files. On our website, the front end is handled by JavaScript, ReactJS, HTML, and CSS. The backend is handled by Solidity. Storage is handled by Ganache as the local host. Finally, this paper will show how to ensure that the procedure is as safe as feasible. We are also maintaining transparency and efficiency here.

Enhancing Cyber Threat Intelligence Sharing through a Privacy-Preserving Federated Learning Approach

This paper proposes a privacy-preserving federated learning approach to enhance cyber threat intelligence sharing. Cyber threats are becoming more sophisticated and are posing serious security risks to organizations. Sharing threat intelligence information can help to detect and mitigate these threats quickly. However, privacy concerns and data protection regulations hinder the sharing of sensitive information. Federated learning is a promising approach that allows multiple parties to collaborate in building a global model while preserving data privacy. We propose a framework that utilizes federated learning to train a global threat intelligence model without compromising the privacy of individual organizations' data. Our approach also includes a differential privacy mechanism to ensure the anonymity of the participating organizations. We demonstrate the effectiveness of our approach through experiments conducted on real-world datasets, showing that it achieves high accuracy while maintaining data privacy. The proposed approach has the potential to facilitate more effective and secure cyber threat intelligence sharing among organizations.

The Anonymity of the Dark Web: A Survey

The dark web is a section of the Internet that is not accessible to search engines and requires an anonymizing browser called Tor. Its hidden network and anonymity pave the way for illegal activities and help cybercriminals to execute well-planned, coordinated, and malicious cyberattacks. Cyber security experts agree that online criminal activities are increasing exponentially, and they are also becoming more rampant and intensified. These illegal cyber activities include various destructive crimes that may target a single person or a whole nation, for example, data breaches, ransomware attacks, black markets, mafias, and terrorist attacks. So, maintaining data privacy and secrecy is the new dilemma of the era. This paper has extensively reviewed various attacks and attack patterns commonly applied in the dark web. We have also classified these attacks in our unique trilogies classification system. Furthermore, a detailed overview of existing threat detection techniques and their limitations is discussed for anonymity providing services like Tor, I2P, and Freenet. Finally, the paper has identified significant weaknesses that make the dark web vulnerable to different attacks.

Leveraging Blockchain Technology Centric Tools in Enhancing the Security Safeguards of Documents Storage

The most widely used method for exchanging data worldwide is the Internet. Numerous cloud service providers support this sharing by letting customers store and share data over the Internet. However, cloud service providers have consistently failed to guarantee data's 100% security regarding privacy. Data breaches, piracy, and hacking attacks have threatened cloud providers' security mechanisms. However, customer data should be 100% secure because it may contain private information that should only be accessible to the owner and an intended audience. As a result, making this system more secure is critical to maintaining data privacy and cloud provider trust. We present a system that incorporates cloud-based data security into the Blockchain. It permits clients to store information in the cloud and gives a specific access control component that guarantees information security. Users can share data with others with permission by sharing the document's link with the intended user. The owner will have access to logs of all operations performed with the document at any time. This will guarantee the data's true ownership and privacy. A third party or individual will be permitted to access the document with prior authorization. This will increase cloud storage's security and reduce data breaches and attacks.

IoT based health monitoring and record management using distributed ledger

Frequent visit of doctors during pandemicis not recommended due to COVID-19 scenario. Internet of Things (IoT) based sensors can be used to measure andto do certain tests at home. The blockchain-based strategy promises to aid in the delivery of health-care services while maintaining data privacy. To maintain transparency in the health record, medications, claim history, etc., up-to-dated distributed ledger is required. Doctors have sometimes necessitate patients to undergo lab tests or buy drugs from those organisations merely in order to generate personal profit, even if the medical shop or pathology labs have a bad reputation. Therefore, blockchain technology provides trust among multiple parties so that any party could not be misguided. At times, the patient visits the hospital in critical condition and they may not be in situation to explain about medical history. In such scenario, the information stored in distributed leger may be helpful to the doctors for further treatment. The medical emergencies such as cardiac arrest and breathing issues, the patient requires immediate treatment or medicine at his/her location, the intelligent drone is useful for faster delivery of medicines and needed medical accessories.

New constructions of equality test scheme for cloud-assisted wireless sensor networks.

Public key encryption with equality test enables the user to determine whether two ciphertexts contain the same information without decryption. Therefore, it may serve as promising cryptographic technique for cloud-assisted wireless sensor networks (CWSNs) to maintain data privacy. In this paper, an efficient RSA with equality test algorithm is proposed. The presented scheme also handles the attackers based on their authorization ability. Precisely, the proposed scheme is proved to be one-way against chosen-ciphertext attack security and indistinguishable against chosen ciphertext attacks. Moreover, the experimental evaluations depict that the underlying scheme is efficient in terms of encryption, decryption, and equality testing. Thus, this scheme may be used as a practical solution in context of CWSNs, where the users may compare two ciphertexts without decryption.

Federated Learning for Microvasculature Segmentation and Diabetic Retinopathy Classification of OCT Data.

PurposeTo evaluate the performance of a federated learning framework for deep neural network-based retinal microvasculature segmentation and referable diabetic retinopathy (RDR) classification using OCT and OCT angiography (OCTA).DesignRetrospective analysis of clinical OCT and OCTA scans of control participants and patients with diabetes.ParticipantsThe 153 OCTA en face images used for microvasculature segmentation were acquired from 4 OCT instruments with fields of view ranging from 2 × 2-mm to 6 × 6-mm. The 700 eyes used for RDR classification consisted of OCTA en face images and structural OCT projections acquired from 2 commercial OCT systems.MethodsOCT angiography images used for microvasculature segmentation were delineated manually and verified by retina experts. Diabetic retinopathy (DR) severity was evaluated by retinal specialists and was condensed into 2 classes: non-RDR and RDR. The federated learning configuration was demonstrated via simulation using 4 clients for microvasculature segmentation and was compared with other collaborative training methods. Subsequently, federated learning was applied over multiple institutions for RDR classification and was compared with models trained and tested on data from the same institution (internal models) and different institutions (external models).Main Outcome MeasuresFor microvasculature segmentation, we measured the accuracy and Dice similarity coefficient (DSC). For severity classification, we measured accuracy, area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve, balanced accuracy, F1 score, sensitivity, and specificity.ResultsFor both applications, federated learning achieved similar performance as internal models. Specifically, for microvasculature segmentation, the federated learning model achieved similar performance (mean DSC across all test sets, 0.793) as models trained on a fully centralized dataset (mean DSC, 0.807). For RDR classification, federated learning achieved a mean AUROC of 0.954 and 0.960; the internal models attained a mean AUROC of 0.956 and 0.973. Similar results are reflected in the other calculated evaluation metrics.ConclusionsFederated learning showed similar results to traditional deep learning in both applications of segmentation and classification, while maintaining data privacy. Evaluation metrics highlight the potential of collaborative learning for increasing domain diversity and the generalizability of models used for the classification of OCT data.

Blockchain and GDPR – A Study on Compatibility Issues of the Distributed Ledger Technology with GDPR Data Processing

This research work aims to investigate Blockchain technology and GDPR compliance studies. This will analyze the data privacy perspective with respect to distributed ledger technology. Blockchain has become one of the most frequently discussed technologies for its ability to allow for peer-to-peer transactions without a centralized intermediary. The GDPR was implemented in May 2018 for EU member states to maintain data privacy. DLT, the underlying technology of blockchain as is a decentralized system without any monetary authority. This research conducted a thorough literature review on prior conducted research to investigate the problems and determine the gaps of GDPR compliance with blockchain technologies and discuss the technical, use-case designs or solutions that make blockchain more compliant GDPR in terms of privacy. This systematic literature review addresses the gaps, feasibility, efficiency, and data privacy issues on compatibility problems that are primarily concerned with how a distributed ledger technology system in which recorded data or transaction cannot be changed or erased is challenging the GDPR data subject access rights (DSAR), where every data subject’s personal data which is compliant to GDPR has a right to exercise their Rights to rectify, delete or limit the processing of your personal data at any time if necessary.

An Efficient Approach Based on Privacy-Preserving Deep Learning for Satellite Image Classification

Satellite images have drawn increasing interest from a wide variety of users, including business and government, ever since their increased usage in important fields ranging from weather, forestry and agriculture to surface changes and biodiversity monitoring. Recent updates in the field have also introduced various deep learning (DL) architectures to satellite imagery as a means of extracting useful information. However, this new approach comes with its own issues, including the fact that many users utilize ready-made cloud services (both public and private) in order to take advantage of built-in DL algorithms and thus avoid the complexity of developing their own DL architectures. However, this presents new challenges to protecting data against unauthorized access, mining and usage of sensitive information extracted from that data. Therefore, new privacy concerns regarding sensitive data in satellite images have arisen. This research proposes an efficient approach that takes advantage of privacy-preserving deep learning (PPDL)-based techniques to address privacy concerns regarding data from satellite images when applying public DL models. In this paper, we proposed a partially homomorphic encryption scheme (a Paillier scheme), which enables processing of confidential information without exposure of the underlying data. Our method achieves robust results when applied to a custom convolutional neural network (CNN) as well as to existing transfer learning methods. The proposed encryption scheme also allows for training CNN models on encrypted data directly, which requires lower computational overhead. Our experiments have been performed on a real-world dataset covering several regions across Saudi Arabia. The results demonstrate that our CNN-based models were able to retain data utility while maintaining data privacy. Security parameters such as correlation coefficient (−0.004), entropy (7.95), energy (0.01), contrast (10.57), number of pixel change rate (4.86), unified average change intensity (33.66), and more are in favor of our proposed encryption scheme. To the best of our knowledge, this research is also one of the first studies that applies PPDL-based techniques to satellite image data in any capacity.

SieveNet: Decoupling activation function neural network for privacy-preserving deep learning

Machine Learning as a Service (MLaaS) is increasingly popular but processing prediction requests will cause users to expose potentially sensitive information. Thanks to breakthrough in secure multi-party computation(MPC), privacy-preserving machine learning has attracted more attention. It allows outsourcing calculations to untrusted servers while maintaining data privacy. Some works proposed frameworks for secure prediction, but with high computation and communication overhead. This is the cost for MPC to deal with non-linear functions, and the activation functions that are widely used in neural networks are usually non-linear.We present a practical framework to perform privacy-preserving prediction. We first propose a new component as an alternative to the activation function so that in one forward propagation, the neural network can be regarded as a linear model. We call this component Sieve Layer and the corresponding network SieveNet. We then show how to use additive secret sharing and adversarial training to build privacy-preserving prediction framework based on SieveNet, and we report a comprehensive analysis of information leakage according to specific types of attacker. Finally, we evaluate our framework on MNIST, CIFAR-10 and CIFAR-100. The results show that the prediction time of our framework is in the same order of magnitude as the plaintext inference.