Non-sensitive Attributes Research Articles

Disentangled contrastive learning for fair graph representations

Graph Neural Networks (GNNs) play a key role in efficiently learning node representations of graph-structured data through message passing, but their predictions are often correlated with sensitive attributes and thus lead to potential discrimination against some groups. Given the increasingly widespread applications of GNNs, solutions are required urgently to prevent algorithmic discrimination associated with GNNs, to protect the rights of vulnerable groups and to build trustworthy artificial intelligence. To learn the fair node representations of graphs, we propose a novel framework, the Fair Disentangled Graph Neural Network (FDGNN). With the proposed FDGNN framework, we enhance data diversity by generating instances that have identical sensitivity values but different adjacency matrices through data augmentation. Additionally, we design a counterfactual augmentation strategy for constructing instances with varying sensitive values while preserving the same adjacency matrices, thereby balancing the distribution of sensitive values across different groups. Subsequently, we employ a disentangled contrastive learning strategy to acquire disentangled representations of non-sensitive attributes such that sensitive information does not affect the prediction of node information. Finally, the learned fair representations of non-sensitive attributes are employed for building a fair predictive model. Extensive experiments on three real-world datasets demonstrate that FDGNN yields the best fairness predictions compared to the baseline methods. Additionally, the results demonstrate the potential of disentanglement in learning fair representations.

Voice Fence Wall: User-optional voice privacy transmission

Sensors are widely applied in the collection of voice data. Since many attributes of voice data are sensitive such as user emotions, identity, raw voice collection may lead serious privacy threat. In the past, traditional feature extraction obtains and encrypts voice features that are then transmitted to upstream servers. In order to avoid sensitive attribute disclosure, it is necessary to separate the sensitive attributes from non-sensitive attributes of voice data. Motivated by this, user-optional privacy transmission framework for voice data (called: Voice Fence Wall) is proposed. Firstly, we provide user-optional, which means users can choose the attributes (sensitive attributes) they want to be protected. Secondly, Voice Fence Wall utilizes minimum mutual information (MI) to reduce the correlation between sensitive and non-sensitive attributes, thereby separating these attributes. Finally, only the separated non-sensitive attributes are transmitted to the upstream server, the quality of voice services is satisfied without leaking sensitive attributes. To verify the reliability and practicability, three voice datasets are used to evaluate the model, the experiments demonstrate that Voice Fence Wall not only effectively separates attributes to resist attribute inference attacks, but also outperforms related work in terms of classification performance. Specifically, our framework achieves 89.84 ​% accuracy in sentiment recognition and 6.01 ​% equal error rate in voice authentication.

On Learning and Testing of Counterfactual Fairness through Data Preprocessing

Machine learning has become more important in real-life decision-making but people are concerned about the ethical problems it may bring when used improperly. Recent work brings the discussion of machine learning fairness into the causal framework and elaborates on the concept of Counterfactual Fairness. In this article, we develop the Fair Learning through dAta Preprocessing (FLAP) algorithm to learn counterfactually fair decisions from biased training data and formalize the conditions where different data preprocessing procedures should be used to guarantee counterfactual fairness. We also show that Counterfactual Fairness is equivalent to the conditional independence of the decisions and the sensitive attributes given the processed nonsensitive attributes, which enables us to detect discrimination in the original decision using the processed data. The performance of our algorithm is illustrated using simulated data and real-world applications. Supplementary materials for this article are available online.

Data Sharing Privacy Metrics Model Based on Information Entropy and Group Privacy Preference

With the development of the mobile internet, service providers obtain data and resources through a large number of terminal user devices. They use private data for business empowerment, which improves the user experience while causing users’ privacy disclosure. Current research ignores the impact of disclosing user non-sensitive attributes under a single scenario of data sharing and lacks consideration of users’ privacy preferences. This paper constructs a data-sharing privacy metrics model based on information entropy and group privacy preferences. Use information theory to model the correlation of the privacy metrics problem, the improved entropy weight algorithm to measure the overall privacy of the data, and the analytic hierarchy process to correct user privacy preferences. Experiments show that this privacy metrics model can better quantify data privacy than conventional methods, provide a reliable evaluation mechanism for privacy security in data sharing and publishing scenarios, and help to enhance data privacy protection.

Sensitive Data Identification and Security Assurance in Cloud and IoT based Networks

Sensitive data identification is a vital strategy in any distributed system. However, in the case of non-appropriate utilization of the system, sensitive data security can be at risk. Therefore, sensitive data identification and its security validation are mandatory. The paper primarily focuses on novel sensitive data recognition methodologies. Further, the sensitivity score of the attributes distinguishes non-sensitive attributes, and domain expert plays an important role in this process. The designing of the security assurance Algo and their corresponding decision tables make the system more robust and reliable. The result section is validated with the help of graphical representation, which clearly makes the authenticity of the research work. In summary, the authors may say that the sensitive data identification and security assurance of the proposed system is automated and work optimally in a cloud-based system.

Achieving data privacy for decision support systems in times of massive data sharing.

The world is suffering from a new pandemic of Covid-19 that is affecting human lives. The collection of records for Covid-19 patients is necessary to tackle that situation. The decision support systems (DSS) are used to gather that records. The researchers access the patient’s data through DSS and perform predictions on the severity and effect of the Covid-19 disease; in contrast, unauthorized users can also access the data for malicious purposes. For that reason, it is a challenging task to protect Covid-19 patient data. In this paper, we proposed a new technique for protecting Covid-19 patients’ data. The proposed model consists of two folds. Firstly, Blowfish encryption uses to encrypt the identity attributes. Secondly, it uses Pseudonymization to mask identity and quasi-attributes, then all the data links with one another, such as the encrypted, masked, sensitive, and non-sensitive attributes. In this way, the data becomes more secure from unauthorized access.

Fair Face Verification by Using Non-Sensitive Soft-Biometric Attributes

Facial recognition has been shown to have different accuracy for different demographic groups. When setting a threshold to achieve a specific False Match Rate (FMR) on a mixed demographic impostor distribution, some demographic groups can experience a significantly worse FMR. To mitigate this, some authors have proposed to use demographic-specific thresholds. However, this can be impractical in an operational scenario, as it would either require users to report their demographic group or the system to predict the demographic group of each user. Both of these options can be deemed controversial since the demographic group is a sensitive attribute. Further, this approach requires listing the possible demographic groups, which can become controversial in itself. We show that a similar mitigation effect can be achieved using non-sensitive predicted soft-biometric attributes. These attributes are based on the appearance of the users (such as hairstyle, accessories, and facial geometry) rather than how the users self-identify. Our experiments use a set of 38 binary non-sensitive attributes from the MAAD-Face dataset. We report results on the Balanced Faces in the Wild dataset, which has a balanced number of identities by race and gender. We compare clustering-based and decision-tree-based strategies for selecting thresholds. We show that the proposed strategies can reduce differential outcomes in intersectional groups twice as effectively as using gender-specific thresholds and, in some cases, are also better than using race-specific thresholds.

A group-key-based sensitive attribute protection in cloud storage using modified random Fibonacci cryptography

Cloud computing is an eminent technology for providing a data storage facility with efficient storage, maintenance, management and remote backups. Hence, user data are shifted from customary storage to cloud storage. In this transfer, the sensitive attributes are also shifted to cloud storage with high-end security. Current security techniques are processed with high encryption time and provide identical security of entire data with single key dependent. These processes are taking high computational time and leaks entire information if the key is hacked. The proposed Group Key Based Attribute Encryption using Modified Random Fibonacci Cryptographic (MRFC) technique rectifies these issues. Instead of machine learning technique, data owner preference-based attributes segregation is used to divide an input dataset into sensitive and non-sensitive attribute groups. Based on inter-organization usage and data owner’s willingness, sensitive attribute is divided into ‘n + 1′ subgroups and each subgroup is encrypted by ‘n + 1’ group keys. The encrypted sensitive subgroups are merged with non-sensitive attributes and uploaded into a private cloud. The novelties of this paper are, (1) data owner preferred sensitive attribute classification instead of machine learning algorithms, (2) sensitive attribute encryption instead of entire attributes, (3) To reduce encryption time without compromising data owner privacy, (4) To decrypt and access the required subgroup instead of the entire attribute. Our experimental results show that, the proposed method takes minimal processing time, better classification accuracy and minimal memory space with high security to selected attributes as compared to existing classification and security techniques. Hence, sensitive data security and privacy is achieved with minimal processing cost.

Blockchain Based Sensitive Attribute Storage and Access Monitoring in Banking System

In earlier banking systems, the data owners were unable to access or update their financial information. Nowadays, user financial information is stored in Cloud storage instead of traditional storage, but security risks are high in financial sectors. Sometimes, Cloud services providers and public auditors modify the user sensitive attribute values. Hence, the decentralized storage system is required for providing better security and integrity of user sensitive attribute values. Blockchain is a decentralized technology for providing tamper-proof storage for sensitive attribute values. Hence, the account holder information is segregated as sensitive and non-sensitive attributes. Before storing into off-chain mode of blocks, the sensitive attributes are grouped into ‘n' number of groups and encrypted by separate group key. The non-encrypted, non-sensitive attributes are stored in cloud storage. The access information is stored in the on-chain mode of the block for easy monitoring. Hence, the security and integrity of sensitive information are preserved.

Relative Privacy Threats and Learning From Anonymized Data

We consider group-based anonymization schemes, a popular approach to data publishing. This approach aims at protecting privacy of the individuals involved in a dataset, by releasing an obfuscated version of the original data, where the exact correspondence between individuals and attribute values is hidden. When publishing data about individuals, one must typically balance the learner ’s utility against the risk posed by an attacker , potentially targeting individuals in the dataset. Accordingly, we propose a unified Bayesian model of group-based schemes and a related MCMC methodology to learn the population parameters from an anonymized table. This allows one to analyze the risk for any individual in the dataset to be linked to a specific sensitive value, when the attacker knows the individual’s nonsensitive attributes, beyond what is implied for the general population. We call this relative threat analysis. Finally, we illustrate the results obtained with the proposed methodology on a real-world dataset.

Sensitivity Based Data Anonymization Model with Mixed Generalization

Published micro-data may contain sensitive information about individuals which should not be revealed. Anonymization approaches have been considered a possible solution to the challenge of preserving privacy while publishing data. Publisheddatasets contain sensitive information. Different sensitive attributes may have different levels of sensitivity. This study presents amodel where the anonymization of tuples is based on the level of sensitivity of the sensory attributes. The study groups sensitiveattributes into highly sensitive and non-sensitive attributes. Tuples with non-sensitive attributes are anonymized. The study conducts experiments with real-life datasets and uses naïve Bayes, C4.5 and simple logistic classifiers to assess the quality of theanonymized dataset. The results from the experiments show that by using the sensitivity based approach to anonymization, thequality of anonymized datasets can be preserved

Critical study and analysis for deciding sensitive and non-sensitive attributes of medical healthcare dataset through survey and using association rule mining

Critical study and analysis for deciding sensitive and non-sensitive attributes of medical healthcare dataset through survey and using association rule mining

A Survey on Antidiscrimination using Direct and Indirect Methods in Data Mining

Data mining is the study of data for relationships that have not previously been discovered. In sociology, discrimination is the hurtful treatment of an individual based on the group, class or category to which that person or things belongs rather than on individual merit: racial and religious intolerance and discrimination. Along with confidentiality, discrimination is a very essential issue when considering the legal and ethical aspects of data mining. It is more than obvious that most people do not want to be discriminated because of their race, gender, religion, nationality, age etc, especially when those attributes are used for making decisions about them like giving them a job, loan, education, insurance etc. Because of this reason, antidiscrimination techniques with discrimination discovery as well as discrimination prevention have been introduced in data mining. Discrimination can be either direct or indirect. Direct discrimination occurs when decisions are taken by considering sensitive attributes. Indirect discrimination occurs when decisions are taken on the basis of nonsensitive attributes which are strongly associated with biased sensitive ones. Here, discrimination prevention in data mining is tackle as well as propose new techniques applicable for direct or indirect discrimination prevention individually or both at the same time. Several decision-making tasks are there which let somebody use themselves to discrimination, such as education, life insurances, loan granting, and staff selection. In many applications, information systems are used for decision-making tasks.

Privacy Preserving Unstructured Big Data Analytics: Issues and Challenges

Big data analytics has created opportunities for researchers to process huge amount of data but created a big threat to privacy of individual. Data processed by big data analytics platforms may have personal information which need to be taken care of when deriving some useful results for research. Existing privacy preserving techniques like, anonymization requires having dataset divided in the set of attributes like, sensitive attributes, quasi identifiers, and non-sensitive attributes. With the structured data it may possible to have such a distribution but in unstructured data it is very difficult to identify sensitive attribute and quasi identifiers.

Discrimination Prevention with Classification and Privacy Preservation in Data mining

Mining of data is an important increasingly methodology for extracting and finding the meaningful hidden knowledge in huge archives of data. There are the various negative social perceptions related to mining of data, out of which many are potential discrimination and potential privacy invasion. The potential discrimination consists of unfairly treating and identifying people based on their existence and belonging to a particular group. Data mining and automated data collection techniques such as classification and association rule mining have provided way to taking decisions automatically, such as computation of insurance premium, loan granting or denial, credit card issue etc. If the provided data sets for training are biased in discriminatory (sensitive) attributes such as, race, gender, religion, etc., discriminatory decisions can be taken and may ensue. For avoiding this situations, antidiscrimination methodology like discrimination prevention and discovery have been considered in the data mining. There are mainly two types of discrimination, one is direct discrimination and second is indirect discrimination. Direct discrimination exists in the situations when decisions are taken on the basis of the sensitive attributes. Indirect discrimination exists in the situations when decisions are taken on the basis of the non-sensitive attributes that are strongly correlated with the biased sensitive attributes.

Collective Data-Sanitization for Preventing Sensitive Information Inference Attacks in Social Networks

Releasing social network data could seriously breach user privacy. User profile and friendship relations are inherently private. Unfortunately, sensitive information may be predicted out of released data through data mining techniques. Therefore, sanitizing network data prior to release is necessary. In this paper, we explore how to launch an inference attack exploiting social networks with a mixture of non-sensitive attributes and social relationships. We map this issue to a collective classification problem and propose a collective inference model. In our model, an attacker utilizes user profile and social relationships in a collective manner to predict sensitive information of related victims in a released social network dataset. To protect against such attacks, we propose a data sanitization method collectively manipulating user profile and friendship relations. Besides sanitizing friendship relations, the proposed method can take advantages of various data-manipulating methods. We show that we can easily reduce adversary's prediction accuracy on sensitive information, while resulting in less accuracy decrease on non-sensitive information towards three social network datasets. This is the first work to employ collective methods involving various data-manipulating methods and social relationships to protect against inference attacks in social networks.

Technique for Prevent Straight and IndirectBias in Information Mining

For coercing the accommodating understanding disguised in the huge accumulation of database the information mining technology is utilized. There are some negative methodologies happened about the data mining technology, among which the potential protection incursion and potential separation. The last comprises of unreasonably treating people on the premise of their having a place with an exact group. Data mining and automated data collection techniques like the arrangement secured the route for making the computerized judgment like granting or denial the loan, race, religion, and so on. On the off chance that the training information set are onesided in what respects oppressive traits like gender, race, religion, and so forth., discriminator decision may guarantee. As a result of this reason the data mining technology presented antidiscrimination strategies with including discrimination discovery and avoidance. The discrimination can immediate or circuitous. At the point when any decisions are made on the sensitive attributes around then direct discrimination are happened. While the indirect discrimination are happened when the decision are made on the premise of non-sensitive attributes which are unequivocally connected with the sensitive. Here in this paper we manage discrimination avoidance in data mining and proposed novel system for discrimination prevention action with the post transforming methodology. We proposed Classification based on predictive association rules (CPAR) algorithm, which is a sort of association classification methods. The calculation joins the preferences of both association classification methods and traditional rule based classification. The algorithm used to avoid discrimination anticipation in post handling. We compute the utility of the proposed approach and contrast and the existing methodologies. The exploratory appraisal demonstrated that the proposed strategy is successfully deleting the direct or indirect discrimination biases in the first data set for keeping up the nature of data.

English

For extracting useful knowledge which is hidden in large set of data, Data mining is a very important technology. There are some negative perceptions about data mining. This perception may contain unfairly treating people who belongs to some specific group. Classification rule mining technique has covered the way for making automatic decisions like loan granting/denial and insurance premium computation etc. These are automated data collection and data mining techniques. According to discrimination attributes if training data sets are biases then discriminatory decisions may ensue. Thus in data mining antidiscrimination techniques with discrimination discovery and prevention are included. It can be direct or indirect. When decisions are made based on sensitive attributes that time the discrimination is indirect. When decisions are made based on nonsensitive attributes which are strongly correlated with biased sensitive ones that time the discrimination is indirect. The proposed system tries to tackle discrimination prevention in data mining. It proposes new improved techniques applicable for direct or indirect discrimination prevention individually or both at the same time. Discussions about how to clean training data sets and outsourced data sets in such a way that direct and/or indirect discriminatory decision rules are converted to legitimate classification rules are done. New metrics to evaluate the utility of the proposed approaches are proposes and comparison of these approaches is also done.

Anonymization on refining partition: Same privacy, more utility

In privacy preserving data publishing, to reduce the correlation loss between sensitive attribute (SA) and non-sensitive attributes(NSAs) caused by anonymization methods (such as generalization, anatomy, slicing and randomization, etc.), the records with same NSAs values should be divided into same blocks to meet the anonymizing demands of ?-diversity. However, there are often many blocks (of the initial partition), in which there are more than ? records with different SA values, and the frequencies of different SA values are uneven. Therefore, anonymization on the initial partition causes more correlation loss. To reduce the correlation loss as far as possible, in this paper, an optimizing model is first proposed. Then according to the optimizing model, the refining partition of the initial partition is generated, and anonymization is applied on the refining partition. Although anonymization on refining partition can be used on top of any existing partitioning method to reduce the correlation loss, we demonstrate that a new partitioning method tailored for refining partition could further improve data utility. An experimental evaluation shows that our approach could efficiently reduce correlation loss.

Classification based on Predictive Association Rule for Discrimination Prevention

In data mining, discrimination is the subject which has been extensively studied in social and economic science. However, there are negative perceptions about data mining. Discrimination comes under two categories one is direct and second is indirect. Decisions based on sensitive attributes are termed as direct discrimination and the decisions which are based on non-sensitive attributes are termed as indirect discrimination which is strongly correlated with biased sensitive once. There are many new techniques proposed for solving discrimination prevention problems by applying direct or indirect discrimination prevention individually or both at the same time. New metrics to evaluate the utility were proposed and are compared with approaches. The proposed work discusses how privacy preservation and prevention between discrimination is implementing with the help of post processing approach. The Classification based on predictive association rules (CPAR) is a kind of association classification methods which combines the advantages of both associative classification and traditional rule-based classification which is used to prevent discrimination prevention in post processing by improving accuracy.