Privacy-preserving Classification Research Articles

An end-to-end framework for private DGA detection as a service.

Domain Generation Algorithms (DGAs) are used by malware to generate pseudorandom domain names to establish communication between infected bots and command and control servers. While DGAs can be detected by machine learning (ML) models with great accuracy, offering DGA detection as a service raises privacy concerns when requiring network administrators to disclose their DNS traffic to the service provider. The main scientific contribution of this paper is to propose the first end-to-end framework for privacy-preserving classification as a service of domain names into DGA (malicious) or non-DGA (benign) domains. Our framework achieves these goals by carefully designed protocols that combine two privacy-enhancing technologies (PETs), namely secure multi-party computation (MPC) and differential privacy (DP). Through MPC, our framework enables an enterprise network administrator to outsource the problem of classifying a DNS (Domain Name System) domain as DGA or non-DGA to an external organization without revealing any information about the domain name. Moreover, the service provider's ML model used for DGA detection is never revealed to the network administrator. Furthermore, by using DP, we also ensure that the classification result cannot be used to learn information about individual entries of the training data. Finally, we leverage post-training float16 quantization of deep learning models in MPC to achieve efficient, secure DGA detection. We demonstrate that by using quantization achieves a significant speed-up, resulting in a 23% to 42% reduction in inference runtime without reducing accuracy using a three party secure computation protocol tolerating one corruption. Previous solutions are not end-to-end private, do not provide differential privacy guarantees for the model's outputs, and assume that model embeddings are publicly known. Our best protocol in terms of accuracy runs in about 0.22s.

Privacy-Preserving Classification on Deep Learning with Exponential Mechanism

How to protect the privacy of training data in deep learning has been the subject of increasing amounts of related research in recent years. Private Aggregation of Teacher Ensembles (PATE) uses transfer learning and differential privacy methods to provide a broadly applicable data privacy framework in deep learning. PATE combines the Laplacian mechanism and the voting method to achieve deep learning privacy classification. However, the Laplacian mechanism may greatly distort the histogram vote counts of each class. This paper proposes a novel exponential mechanism with PATE to ensure the privacy protection. This proposed method improves the protection effect and accuracy through the screening algorithm and uses the differential privacy combination theorems to reduce the total privacy budget. The data-dependent analysis demonstrates that the exponential mechanism outperforms the original Laplace mechanism. Experimental results show that the proposed method can train models with improved accuracy while requiring a smaller privacy budget when compared to the original Pate framework.

EV-Perturb: event-stream perturbation for privacy-preserving classification with dynamic vision sensors

EV-Perturb: event-stream perturbation for privacy-preserving classification with dynamic vision sensors

A hybrid data perturbation and mean clustering approach based privacy preserving classification model for large databases

A hybrid data perturbation and mean clustering approach based privacy preserving classification model for large databases

Privacy-Preserving Classification Scheme Based on Support Vector Machine

As a classifier, support vector machine (SVM) explains a core problem of machine learning, namely sample classification in statistical terms. It has been widely used in machine learning, data mining, pattern recognition, and other fields. With the wide applications of SVMs in machine learning and big data, privacy protection of sensitive data in SVMs is becoming more and more important, such as face recognition and biometric information. At present, the main privacy protection methods in SVMs are homomorphic encryption and secure multiparty computation. However, there are some problems in current research. The computational efficiency is low, and the scalability of the schemes is poor. In addition, the user must stay online in some solutions. To solve the abovementioned problems, this article designs a secure and efficient classification scheme based on SVM to protect the privacy of private data and support vectors in the calculation and transmission process. First, the distributed two trapdoors public-key cryptosystem proposed by Liu is used to realize the distributed double-key decryption function, weaken the decryption capability of a cloud server with the master key, and prevent the server from launching active attacks. Second, we design a universal secure computing protocol for nonlinear SVMs based on the Gaussian kernel function, which can be extended to polynomial kernel function, sigmoid kernel function and is suitable to different kernel functions. Compared with existing schemes, our solution reduces the amount of encrypted data, simplifies the calculation process, and improves calculation efficiency. Third, an introduced cloud server realizes user offline function. Finally, we analyze the security of the scheme and verify its efficiency through experiments. Analysis and experimental results show that the scheme has the advantages of high efficiency, good scalability, and user offline function.

A novel cryptographic protocol for privacy preserving classification over distributed encrypted databases

A novel cryptographic protocol for privacy preserving classification over distributed encrypted databases

A Secure Clinical Diagnosis With Privacy-Preserving Multiclass Support Vector Machine in Clouds

With the rapid growth of machine learning and artificial intelligence in medical cloud systems, cloud-aided medical computing has materialized a concrete platform for rapid realization of the service-oriented computing paradigm. However, using machine learning approaches in medical clouds might cause serious privacy leakage risk, for example, it needs to take the patients’ sensitive data as the inputs of the machine learning algorithm. In this article, we propose an efficient and privacy-preserving clinical diagnosis scheme that was performed by an (untrusted and malicious) outsourced cloud platform, which can provide the diagnosis assistance for doctors without leaking any sensitive information of patients and service providers. Concretely, we give the security model of privacy-preserving multiclass support vector machine (SVM) in outsourced medical clouds and design a secure clinical diagnosis scheme with privacy-preserving multiclass predecision and diagnosis. We propose a novel encoding approach to implement the encryption of negative number, and design several security building blocks, such as privacy-preserving decision function computing, privacy-preserving classification, and search of max decision function on encrypted fields, to provide the construction of the secure clinical diagnosis with privacy-preserving multiclass SVM. We give the concrete scheme and provide the experiment on Dermatology datasheet. Security analysis and experimental results indicate that the proposed scheme is efficient and practical in privacy-preserving clinical diagnosis systems.

Privacy-Preserving Classification in Multiple clouds eHealthcare

Internet of Things (IoT) is increasingly being used in real life, especially in the eHealthcare field. Among eHealthcare, the application of predicting patients' health status based on their daily activity data which is collected by IoT equipment has attracted extensive attentions and researches. In this application, patients' data which are treated as time-series data are transmitted to healthcare center (HC), then HC makes predictions based on an established classification model. However, making predictions using classification models requires a lot of computing resources, while HC usually cannot afford such numerous calculations. The use of the cloud solves the problem of insufficient computing resources, but it causes another problem, namely the leakage of user privacy. In particular, not only patients' data leak patients' privacy information, the classification model also causes the privacy disclosure of patients and HC. We design a new system model and propose an algorithm which can protect patients' data and classification model from leakage and offload calculation to multiple clouds. Our algorithm can better protect privacy of patients and HC in more complex classification scene, and can effectively reduce the computational cost of the healthcare center

A differential privacy-based classification system for edge computing in IoT

The blooming Internet of Things (IoTs) devices have brought many new types of sensing applications and methods to the traditional cloud-enabled IoTs framework. Hence, the new network framework becomes bidirectional gradually, in which IoTs devices can also perform moderate computation tasks instead of being solely used as data harvesters. This new coming framework known as edge computing significantly improves the traditional cloud-enabled network latency and dependency by shifting part of computation back to “local”. However, new security risks emerge when the edge computing shifts data and models back to the IoTs devices.Acies, a differential privacy based privacy-preserving classification system for edge computing is proposed to secure the classification models offloaded to edge devices. Acies supports popular classifiers such as Nearest Neighborhood, Support Vector Machine and Sparse Representation Classifier with a variety of feature selection methods. According to our evaluation on different datasets, classification models with Acies can be private and maintain high utility. Acies achieves reliable privacy protection under reconstruction attacks with minimal impact on classification accuracy (2% ∽ 5%) only. Acies outperforms the naive input dataset perturbation methods by up to 30% higher classification accuracy when the privacy requirements of the applications is high (the privacy budget ε is less than 2).

Privacy Preserving Classification of EEG Data Using Machine Learning and Homomorphic Encryption

Data privacy is a major concern when accessing and processing sensitive medical data. A promising approach among privacy-preserving techniques is homomorphic encryption (HE), which allows for computations to be performed on encrypted data. Currently, HE still faces practical limitations related to high computational complexity, noise accumulation, and sole applicability the at bit or small integer values level. We propose herein an encoding method that enables typical HE schemes to operate on real-valued numbers of arbitrary precision and size. The approach is evaluated on two real-world scenarios relying on EEG signals: seizure detection and prediction of predisposition to alcoholism. A supervised machine learning-based approach is formulated, and training is performed using a direct (non-iterative) fitting method that requires a fixed and deterministic number of steps. Experiments on synthetic data of varying size and complexity are performed to determine the impact on runtime and error accumulation. The computational time for training the models increases but remains manageable, while the inference time remains in the order of milliseconds. The prediction performance of the models operating on encoded and encrypted data is comparable to that of standard models operating on plaintext data.

Privacy preserving rule-based classifier using modified artificial bee colony algorithm

Privacy preserving data mining is a hot research field of data mining. The aim of privacy preserving data mining is to prevent the leakage of the sensitive information of individuals while performing data mining techniques. Classification task is one of the most studied fields in data mining hence in privacy preserving data mining as well. On the other hand, differential privacy is a powerful privacy guarantee that determines privacy leakage ratio by using ∊ parameter and enables researchers to mine data which includes sensitive information. Implementations of some well-known classification algorithms such as k-NN, Naïve Bayes, ID3, etc. with differential privacy have been developed. Although the success of the rule-based classifiers using meta-heuristics such as Ant-Miner, BeeMiner etc. in data mining has been demonstrated, any implementation of these classification algorithms with differential privacy has not been proposed in the literature until now to our best knowledge. Artificial bee colony (ABC) is a nature inspired algorithm which imitates foraging behavior of bees, and some approaches using ABC to discover classification rules have been proposed recently and the success of ABC algorithm for the discovery of classification rules has been demonstrated. Motivated by this shortcoming in the literature, we propose to develop a rule-based classifier using ABC algorithm with input perturbation technique of differential privacy to perform privacy preserving classification. According to our experimental results, the proposed ABC-based classifier performs better than the well-known algorithms that are SVM, C4.5, Holte’s One Rule, PART, and RIPPER over non-private and differentially private versions of the datasets in terms of classification performance.

Efficient Model for Privacy Preserving Classification Of Data Streams

Privacy preserving data mining has become progressively mainstream since it permits sharing of privacy delicate data for examination purposes .So individuals have gotten progressively reluctant to share their data, regularly bringing about people either declining to share their data or giving inaccurate data. As of late, privacy preserving data mining has been concentrated broadly, on account of the wide multiplication of delicate data on the web. Data Mining manages programmed extraction of already obscure examples from a lot of data sets. These data sets ordinarily contain touchy individual data or basic business data, which thusly get presented to different gatherings during Data Mining exercises. This makes hindrance in Data Mining measure. Answer for this issue is given by Privacy preserving in data mining (PPDM). PPDM is a specific arrangement of Data Mining exercises where procedures are developed to secure privacy of the data, so the information revelation cycle can be completed without obstruction. The target of PPDM is to shield delicate data from spilling in the mining cycle alongside exact Data Mining results. The objective of this paper is to introduce the survey on different privacy preserving strategies which are useful in mining huge measure of data with sensible productivity and security.

Novel Privacy Preserving Classification Mining Approach Applied to A City Public Security Big Data Analysis

With the development of the big data analysis technology, more and more companies are willing to apply this technology. In order to further promote the big data investigation teaching of Sichuan Police College, we accurately grasp the new situation, new tasks and new requirements faced by the big data investigation, and further deepen the solidification of the implementation path. Through on-the-spot visit, discussion and communication, we will focus on understanding the application technology of big data, extend it to various work of public security organs’ attack, prevention and control. Then we want to promote the establishment of practical application systems, and promote the fundamental change of criminal investigation mode through the construction of big data.

Quantifying Membership Privacy via Information Leakage

Machine learning models are known to memorize the unique properties of individual data points in a training set. This memorization capability can be exploited by several types of attacks to infer information about the training data, most notably, membership inference attacks. In this paper, we propose an approach based on information leakage for guaranteeing membership privacy. Specifically, we propose to use a conditional form of the notion of maximal leakage to quantify the information leaking about individual data entries in a dataset, i.e., the entrywise information leakage. We apply our privacy analysis to the Private Aggregation of Teacher Ensembles (PATE) framework for privacy-preserving classification of sensitive data and prove that the entrywise information leakage of its aggregation mechanism is Schur-concave when the injected noise has a log-concave probability density. The Schur-concavity of this leakage implies that increased consensus among teachers in labeling a query reduces its associated privacy cost. Finally, we derive upper bounds on the entrywise information leakage when the aggregation mechanism uses Laplace distributed noise.

A hybrid approach of homomorphic encryption and differential privacy for privacy preserving classification

Privacy preserving data mining is a substantial research area that aims at protecting the privacy of individuals while enabling to perform data mining techniques. In this study, we propose a secure protocol that fulfils the privacy restriction by combining homomorphic encryption with differential privacy and integrate this protocol into Holte’s One Rule which is a simple, but accurate and efficient classification algorithm. The proposed method allows a researcher to get the answers of his/her queries to build One Rule classifier by processing the encrypted training dataset under Paillier’s cryptosystem and also applies differential privacy to minimize the privacy leakage of individuals as much as possible in this training dataset. Therefore, both of security and privacy of the individuals in the training dataset for classification are provided thanks to our proposed method; since neither the parties, nor the researcher attain any information about the individuals in the database. Besides the One Rule classifier, we apply our proposed privacy preservation model to Naïve Bayes classification algorithm for the performance comparison, and show the efficiency of the proposed method through experiments on real data from UCI repository.

An Improved Rotation-Based Privacy Preserving Classification in Web Mining Using Naïve Bayes Classifier

An Improved Rotation-Based Privacy Preserving Classification in Web Mining Using Naïve Bayes Classifier

An Improved Rotation-Based Privacy Preserving Classification in Web Mining Using Naïve Bayes Classifier

An Improved Rotation-Based Privacy Preserving Classification in Web Mining Using Naïve Bayes Classifier

Privacy preserving classification over differentially private data

AbstractPrivacy preserving data classification is an important research area in data mining field. The goal of a privacy preserving classification algorithm is to protect the sensitive information as much as possible, while providing satisfactory classification accuracy. Differential privacy is a strong privacy guarantee that enables privacy of sensitive data stored in a database by determining the ratio of sensitive information leakage with respect to an ɛ parameter. In this study, our aim is to investigate the classification performance of the state‐of‐the‐art classification algorithms such as C4.5, Naïve Bayes, One Rule, Bayesian Networks, PART, Ripper, K*, IBk, and Random tree for performing privacy preserving classification. To preserve privacy of the data to be classified, we applied input perturbation technique coming from differential privacy, and observed the relationship between the ɛ parameter values and accuracy of the classifiers. To our best knowledge, this article is the first study that analyzes the performances of the well‐known classification algorithms over differentially private data, and discovers which datasets are more suitable for privacy preserving classification when input perturbation is applied to provide data privacy. The classification algorithms are compared by using the differentially private versions of the well‐known datasets from the UCI repository. According to the experimental results, we observed that, as ɛ parameter value increases, better classification accuracies are achieved with lower privacy levels. When the classifiers are compared, Naïve Bayes classifier is the most successful method. The ɛ parameter should be greater than or equal to 2 (i.e., ɛ ≥2) to achieve cloud server is malicious and untrusted, sensitive data will satisfactory classification accuracies.This article is categorized under: Commercial, Legal, and Ethical Issues > Security and Privacy Technologies > Classification

Differentially private 1R classification algorithm using artificial bee colony and differential evolution

Classification is an important topic in data mining field. Privacy preserving classification is a substantial subtopic that aims to perform classification of private data with satisfactory accuracy, while allowing sensitive information leakage at minimal level. Differential privacy is a strong privacy guarantee that determines privacy leakage ratio by using ϵ parameter; and enables privacy of individuals whose sensitive data are stored in a database. There exist some differentially private implementations of well-known classification algorithms such as ID3, random tree, random forests, Naïve Bayes, SVM, logistic regression, k-NN etc. Although One Rule (1R) is a simple but powerful classification algorithm, any implementation of differentially private 1R classification algorithm has not been proposed in the literature to our best knowledge. Motivated by this gap, first we propose a differentially private 1R classification algorithm (DP1R), then improve its performance by using metaheuristics that are differential evolution (DE) and artificial bee colony (ABC) in this study. Additionally, we also apply DE and ABC to improve performance of differentially private Naïve Bayes classifier and compare with DP1R. Moreover, DP1R is compared with the state-of-the-art differentially private algorithms such as differentially private SVM, differentially private logistic regression, differentially private ID3, and differentially private random tree on nine publicly available UCI datasets. The experimental results demonstrate that DP1R is an efficient classifier that has very similar accuracy to differentially private SVM which has the best accuracy results, however with respect to running time comparison of the methods, DP1R has the best performance among the all methods.

Techniques and Challenges while Applying Machine Learning Algorithms in Privacy Preserving Fashion

Nowadays many different entities collect data of the same nature, but in slightly different environments. In this sense different hospitals collect data about their patients’ symptoms and corresponding disease diagnoses, different banks collect transactions of their customers’ bank accounts, multiple cyber-security companies collect data about log files and corresponding attacks, etc. It is shown that if those different entities would merge their privately collected data in a single dataset and use it to train a machine learning (ML) model, they often end up with a trained model that outperforms the human experts of the corresponding fields in terms of accurate predictions. However, there is a drawback. Due to privacy concerns, empowered by laws and ethical reasons, no entity is willing to share with others their privately collected data. The same problem appears during the classification case over an already trained ML model. On one hand, a user that has an unclassified query (record), doesn’t want to share with the server that owns the trained model neither the content of the query (which might contain private data such as credit card number, IP address, etc.), nor the final prediction (classification) of the query. On the other hand, the owner of the trained model doesn’t want to leak any parameter of the trained model to the user. In order to overcome those shortcomings, several cryptographic and probabilistic techniques have been proposed during the last few years to enable both privacy preserving training and privacy preserving classification schemes. Some of them include anonymization and k-anonymity, differential privacy, secure multiparty computation (MPC), federated learning, Private Information Retrieval (PIR), Oblivious Transfer (OT), garbled circuits and/or homomorphic encryption, to name a few. Theoretical analyses and experimental results show that the current privacy preserving schemes are suitable for real-case deployment, while the accuracy of most of them differ little or not at all with the schemes that work in non-privacy preserving fashion.