|Intelligent Automation & Soft Computing
Multi Attribute Case Based Privacy-preserving for Healthcare Transactional Data Using Cryptography
Department of Computer Science and Engineering, Bannari Amman Institute of Technology, Sathyamangalam, 638401, India
*Corresponding Author: K. Saranya. Email: email@example.com
Received: 29 January 2022; Accepted: 07 March 2022
Keywords: Privacy-preserving; crypto policy; medical data mining; integrity and verification; personalized records; cryptography
Personal health information system is an important paradigm in healthcare industry that collects and stores the patient data including sensitive information like demographic data, medical history, diagnostic codes, treatment plans, lab records, insurance information, immune dates, allergies, etc. . Privacy preservation in medical healthcare data has gained much attention in the recent years, especially information mining analysis in big data analysis play an important role for healthcare decisions. Healthcare industry generates huge volumes of medical data which should be analyzed with caution, as the number of patterns to be processed are high. Further, this healthcare data contains both sensitive and non-sensitive information in both structured and non-structured format. So, information access during any transactions pose severe threats to individual’s data security in health care industry. Most of the patients have records, directed from hospitals, and it contains their Personal Health Information (PHI) which are used before treating a person. In this context, Personal Health Information (PHI) of a patient is shared among different departments of a hospital namely, pharmacy, lab, consulting, admin etc. Such medical records contain sensitive information with regard to the patients’ medical history. The chances for this information to be misused by anonymous intruders are extremely high. Private information leakage issue leads to information theft from patient’s records or any other healthcare related documents. A patient’s privacy and data security is breached when their healthcare data is shared without proper channel. So there is a need exists to develop a viable way that balances both healthcare data sharing as well as privacy protection . In this study, a new privacy preserving framework is to be designed in the form of attribute-based protection using crypto analysis. This framework is developed for healthcare conditions and manages raw transaction datasets. A multi-attribute based protection policy is proposed using crypto policy standard to protect the raw data from segregating sensitive attributes which restrict unauthorized access and gives reverse hidden crypto policy to protect the sensitive information. This leads to the protection of sensitive medical information which is an important milestone in privacy preserving data policy.
Few privacy-preserving methods exist with an aim to protect the appropriate information of patients, while at the same time, it also maintains their private information including healthcare details. To accomplish this aim, a large number of privacy-preserving models is proposed to hide the sensitive information from public setting. Such Privacy-Preserving Data Publishing (PPDP) models are utilized to secure different types of information including socioeconomics and analysis codes with an expectation to keep the dangers of character disclosures such as trait disc1osure and enrollment disclosure as shown in Fig. 1. Further, the privacy methods have evolved with another meaning of privacy called ‘k secrecy’ which roughly translates into verified access from utility access logs. Being a namelessness demonstrate, it provides structure to the calculation whereas the frameworks disseminate the information, after anonymization. It secures the information that correlates to a specific substance. It principally keeps the private data away from linkage assault. Namelessness is accomplished by leveraging both concealment and speculation.
Most healthcare industry processes directly make use of data about patient’s healthcare information without any protection. The industries do not concentrate on the influence of attacks upon sensitive information. In this scenario, outsider access becomes sophisticated and data access becomes easy. Otherwise, encryption policy is commonly employed to unhide or hide the whole data. The attribute case is non-violent against healthcare sensitive information and so the access restriction becomes an important element to verify the log of authentication.
When the personalized medical information is separated as attributes in the form of sensitive and non-sensitive patient’s records, it specifies the existence of highly confidential and classified sensitive information. Medical reports contain the personal information of patients in unstructured formats. It contains critical information of the patient and their caretakers such as names, telephone numbers, locality, age and furthermore the points of sicknesses.
The medical records contain the original information, directed from hospital records, based on the diagnosed information. So, the data records must be open access for anyone to read the information. The sensitive information, which are non-secure in nature, contains various types of personalized information. The information from patients include general details of the patient like name, phone number and address while these data are not sensitive in nature than the treatment and disease facts.
For example, DNA sample and drugs intake levels are sensitive information in a case record of a patient. This medical record includes the patient’s original drug details and if it is accessed by the hackers through unreasonable approaches, they may exploit the data to enhance their business interests and continue to do so. However, -crypto based hidden policy hides the sensitive information excluding authorized person who possess rights to access and ensure key authentication to access the data.
With the implementation of the proposed crypto policy, the patient’s information logs in healthcare industry get high security. This can be achieved by improving privacy preserving framework in the form of provable privacy security policy. During the implementation of this policy, verifiable resources are verified which contain attributes that are highly sensitive and insensitive information that does not affect the transactions done in healthcare domain, by maintaining the privacy of the patient. Masking identifiers specify the sensitive information of suffix hidden crypto encryption standard that maintains the individual records of the patient.
2 Literature Survey
Privacy preserving approach is an important technique to secure the personalized data and privacy information of individuals. This is applicable especially in case of healthcare details of the patients . Though most of the privacy preserving techniques secure personalized information and are difficult to use the differential security to control information, these techniques fail in hiding private data as sensitive information. So healthcare data and the information of patients get distributed freely. This sensitive data gets uncovered which is an essential issue to be overcome. One of the data privacy protection techniques called k-namelessness is implemented based on crypto policy. In k-anonym zed dataset , each record is undefined from any rate whereas k −1 possesses different records regarding certain unique traits that depend upon the hiding principle.
Anonymization of restorative records is of extraordinary significance in healthcare domain, since uniform content can be easily accessible. This data about personal information of patients, accessed from healthcare specialist’s records, can be protected based on Machine Learning approach named Principle Component Analysis (PCA) . This is inclusive of bouncing the vulnerability of the calculated relapse and annoying the scholarly classifier with clamor that corresponds to the vulnerability . When considering this approach, there are constraints present in it, when connected with other machine learning approaches. At that point, another privacy protection robust reversible calculation algorithm is discussed . The proposed Scheme Against Global Eavesdropping (SAGE) can accomplish both security and provide relevant protection against global attackers. Privacy Preserving Data Publishing (PPDP) has a lot of strategies and devices to disseminate valuable data while safeguarding the information. The novel study is conducted to reveal how regression and correlation analysis can be performed on vertically-divided data using Locality Sensitive Hashing (LSH) approach . As of late, PPDP has got impressive track record to look into groups, and numerous methodologies have been proposed for different types of information dissemination scenarios . This sort of prerequisite is considerably harder to be addressed for dynamic and expansive scale joint efforts. Here, the quantity of access control approaches are huge in nature for both dissemination and exploration of the data . In literature, Internet of Things (IoT) has been proposed to reconcile the activities that assess the application of strategies in health care applications .
In literature, the researchers investigated the necessities of a correspondence structure brought together from different individual entities. This study broke down the storyline based on a day in Robert’s life. In the study conducted earlier , Secure Personal health information Sharing (SPS) was proposed which envelops personality-based cryptography to guarantee security and privacy of PHI by utilizing short advanced signature and pseudo-identity of the patient . SPS reduces the burden of Health Service Provider (HSP) in terms of PHI stockpiling and administration. Further, it also supports the same by joining distributed storage administrations to electronic Health (eHealth) mind framework . PHR is a sort of wellbeing record recorded and maintained by both patient and the healthcare professionals. A perfectly-maintained individual healthcare record can provide therapeutic data of the patient from various sources and help in decision making for the patient’s wellbeing [15,16]. The medicinal synopsis from PHR can be accessed through Internet or other such media when there is a necessity for security and privacy.
From healthcare data security perspective, healthcare data is processed through Mobile Healthcare Networks (MHN) which needs support so that the quality of access verification can be enhanced. Further, a few countermeasures can also be taken for security and privacy assurance in MHNs including privacy-preserving healthcare information integration , preparation of secure healthcare information, and troubles in recognizing the information. Fine-Grained and versatile information reach the control plot in light of Attribute Based Encryption (ABE). Furthermore, PHI sharing approaches might be sensitive  and may uncover data about hidden PHI or information senders or recipients . In current study, each characteristic is allowed to have a property name along with pride. Further, it is allowed to embrace the Bloom channel to check the traits before decryption efficiently. A certifiable and unavoidable health monitoring application (inescapable fall recognition for stroke alleviation) is developed to show the adequacy and viability of haze registered across the globe in wellbeing monitoring . The author proposes an Adaptive Private Security (APS) algorithm to protect the sensitive information in healthcare industry [21,22].
3 Multi Attribute Case-Based Privacy Preservation Scheme
The healthcare processes contain sufficient security resource to carry out the protection of personalized data so as to safeguard the privacy information of patients. Conventional medical data is huge in quantity while the personal information cannot be easily mined out by product developers. Hence, the current study proposes to privacy-preserving the sensitive data from patent logs using Multi Attribute Case Based Privacy Preservation (MACPP). This is proposed for the purpose of improving the security of patient data. Healthcare information is viewed as an extremely critical aspect in today’s world. Such data must be handled in a sensitive manner using advanced techniques since the patient’s personal information are at risk. When such data is distributed without any inhibition, the data becomes defenseless against any types of attacks. Numerous methods have been proposed earlier to safeguard the privacy of healthcare information. In current research work, a review of the models and systems proposed earlier is shown while these models are used for distributing information about patients.
In this method, preprocessing is the first step to clean the raw data from medical dataset collection. Each transaction from the medical records contains different types of sensitive information in terms of drugs. So, this healthcare information should be accessed both on a private and general level. But the usage of personal medical information is different from general information of the patient. This should be understood finely to distinguish the sensitive information and protect it through hidden policy. Such sensitive data should be represented on raw conditions themselves which need privacy protection to be accessed by unauthorized persons.
The personal data collected from patients are always in raw state and can be segregated based on sensitive attributes from patient’s records. Based on these values, the mean value of average is determined for both descriptive sensitive and non-sensitive data as shown in Fig. 2. The average counts find the weightage factors for highly sensitive drug usage of the patients to use service level encryption standard that can hide the data. Finally, the linear integrity verifies the privacy terms against a user’s key validation to prove the security.
• Security and Privacy in Data Mining
Privacy is regularly characterized as a condition in which the sensitive information is secured by any means. It focuses on the utilization and administration of a person’s personal information like making arrangements and setting up approval prerequisites so that the patient’s personal information is gathered, shared and used in appropriate ways. By improving the security principles, the information can be partitioned in which the personalized information gets different access from general information that can be accessed by those only who has authorization. It focuses on shielding the information for authentication processes and taking information for a benefit by an authenticated person. In spite of the fact that security is imperative to ensure the information is safe, the privacy still lacks.
• Objective consideration
From the proposal, it can be considered to apply data separation based on provable partition crypto techniques so that privacy can be preserved in classifying medical data. The authors considered two approaches to ensure privacy such as vertical partition and horizontal partition. In vertical partition approach, each site utilizes a segment of credits to register its outcomes. Followed by, the appropriate outcomes are collected at a focal point which gathers huge amount of outfit strategy. In horizontal partition approach, the information is conveyed among a few locales. Each site processes a particular information and a focal trust gathers information about these outcomes.
3.1 Preprocessing Medical Datasets
Across the healthcare institutions, medical data sets are collected in the form of raw data so as to maintain the patient log. The information collected in hospitals form records with attributes that contain patient’s information, drugs, suggestions and consent forms. The unavoidable yet unnecessary information is called as noise since it interrupts the information that provides much insights about the patient’s health history. The attributes are nonlinear in its raw stages since it is yet to be pre-processed. Since empty fields bring noise to the analysis, record omission is carried out so that the important records are not avoided. Further, distinct outliers from medical data records are removed.
Tab. 1 demonstrates the preprocessed dataset out of raw pharm medical. This dataset contains information on patient id, name, age, mobile number, diagnose and drugs name & dose.
The above algorithm clears the raw data (Rs), which was collected as data without any outliers. to form distinct values from sensitive and non-sensitive information. Preprocessing narrows the raw data from a conventional approach in which each record is considered as an attribute while empty case are taken as Null field to reduce the dimension and simplify the privacy principle process.
3.2 Multi Attribute Case Reasoning
The attributes are measured to ensure data dependency and segregate the transactional data as sensitive and non-sensitive attributes. The verified category of sensitive case attributes is different from standard attribute case, for example, the paramedical counts from normal patients differ from other high-risk patients. So, the physicians prescribe drugs with highly sensitive information to treat the patients. In this point of view, privacy concerns arise due to the involvement of high sensitive data. This considers multiple attributes based on the relation between drug and diagnosed disease condition, and compound molecule pattern to process the reasoning principle and protect the data. The provable partition differentiates the sensitive attributes with differential count measure of other drug average mean values.
The raw attribute is obtained from a confidential list which has a cross mean value of drug representation to combine the records as shown in Tab. 2.
The above algorithm splits the attributes into sensitive (Si) and non-sensitive (Nsi) based on results of relative mean analyses obtained from sensitive factors.
3.3 Hidden Crypto Security Analysis
In this stage, privacy principles verify the nature of outsourced data based on access principle so that cryptographic principle can be applied. This makes the sensitive information safer and its fortunes of privacy information cannot be misused. So, the suffix patterns are hidden to encrypt the data. In healthcare industry, patient’s private information is highly important since it involves drugs and prescriptions that are hidden based on specified principles and access rights. The hidden crypto policy encrypts the sensitive information from logical suffix representation of each attribute. In records, it can be reversed to hide the information with specific characters using symmetric key verification standards.
The sensitive attribute enriches the least maximum of suffix value and secures the information from unauthorized access. Tab. 3 shows the suffix hidden crypto policy for personal health information which is divided into sensitive and non-sensitive attributes.
3.4 Linear Integrity for Verifiable Outsource Data
Linear integrity verifies the attribute case records which are sensitive in nature in terms of access rights on verification and validation against hackers. These hackers attempt to encrypt and decrypt the information on symmetric standard and hold the information constructs based on confirmation. Further, the trustworthiness of such data needs to be checked and the outsourced data must be reworked to meet the customer’s demand. In addition, secure data sensitivity approach is proposed with enhanced security. This plan too helps in gaining extensive evidence and ensure dynamic skill activity simultaneuosly.
4 Result and Discussion
Multi attribute case-based privacy preservation algorithm was proposed and implemented in this study to assess its effectiveness upon patients’ health care records sourced from UCI web repository. The proposed privacy principle implemented a context-aware multi attribute relational analysis with hidden crypto model to deduce a set of privacy fields from a user’s Electronic Health Records (EHR). The proposed method produced efficient results on context-aware clustering and improved the performance too. Principle Component Analysis (PCA), Locality Sensitive Hashing (LSH) and Adaptive Private Security (APS) are some of the algorithms used for comparison and its parameters are tabulated herewith.
Tab. 4 shows the details of data set used to evaluate the performance of the proposed Multi-Attribute Provable Partition-based approach. The performance of MAPPC was evaluated through privacy accuracy (cs), precision rate, recall rate and time complexity.
In Fig. 3, the comparison results for privacy-preserving accuracy are shown which infer that the proposed method produced the highest performance than other methods.
Tab. 5 shows the comparison results of privacy preserving accuracy achieved by the proposed method against existing techniques. The proposed model achieved privacy preserving accuracy values such as 96.8%, 97.9%, and 98.4% under 1000, 2000, and 3000 records respectively.
4.1 Analysis of Precision Rate
Precision (Pr) is defined as the proportion of total number of relevant sensitive information from healthcare dataset, where R is the relevant margin rate calculated and A is the total number of sensitive data.
Fig. 4 shows the comparison results for precision rate produced by the proposed method against existing methods. The results infer that the proposed method produced an excellent performance rate than other methods.
Tab. 6 shows the results accomplished by the proposed approach i.e., high performance ratios such as 89.1%, 85.4%, and 86.8% under 1000–3000 records respectively.
4.2 Analysis of Recall
Recall (Rc) is defined as the proportion of matched drug items within relevant drugs against the total drugs with high sensitivity.
Fig. 5 shows the comparison results of false recall ratio produced by the proposed method against other methods. The proposed method achieved excellent performance over other methods.
Fig. 6 shows the comparison results of time complexity achieved by both the proposed method and other methods considered for the study. The proposed approach produced less time complexity compared to other methods.
Tab. 7 shows the comparison results of recall value which infer that the proposed method accomplished the highest performance compared to other methods.
4.3 Analysis of Time Complexity
Tab. 8 shows the comparison of time complexity of the proposed precision clustering which offered time complexity values such as 4.7, 5.2 and 7.1 ms under 1000–3000 records respectively.
Healthcare privacy techniques focus on sensitive information collected from patient logs such as drug information, patient’s medical history. So, such sensitive information must be protected from intruders. The current study proposed a multi attribute sensitive protection method using relational analyses using benchmark healthcare data. The proposed MACPP considered radical patient-centric information from drugs category with sensitive personal prescription in order to achieve provable linear partition and enhance the level of security. The performance of the proposed framework, in terms of accuracy, was found to be higher compared to other methods. The proposed method improved the performance of privacy accuracy by 96.8%, with a least time complexity of 4.7 milliseconds. The experimental results established the privacy-preserving ability of state-of-the-art model which is phenomenal and better than existing methods. In future, hybrid deep learning models can be employed to further improve the security of sensitive data.
Funding Statement: The authors received no specific funding for this study.
Conflicts of Interest: The authors declare that they have no conflicts of interest to report regarding the present study.
|This work is licensed under a Creative Commons Attribution 4.0 International License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.