An Efficient Privacy Preserving Approach for e-Health

Immense Procreation of large amounts of data in medical field and health care domain, benefitting society is at risk with sensitive attributes being disclosed. Access to Medical Information made feasible over internet with an intension of serving the people related to medical community is triggering a challenge for researchers in norms of Privacy and security. The medical data at cloud is vulnerable to unpredictable threats with evolving technology, and the threat landscape sounds resilient with sensitive attributes. In this contemporary stretch, Organizations fail to hold the reputation and are unable to preserve public confidence. The austerity of sophisticated security attacks compromise the privacy of patient data and security of healthcare units. The fruitful approaches by several researches and practitioners provided an up heal resolutions, but the demand for an optimal solution remains unanswered. In this paper we present a solution for addressing the security issues in health care management. We propose a hybrid framework using enhanced Attribute Based Encryption (ABE) with Anonymity approach based on access primitives of sensitive attributes. The proposed mechanism is evaluated in terms of performance, encryption time, decryption time and Memory utilization using Jsim simulator which envisage drastic performance expedition in the presented model. Keywords—e-Health; Attribute Based Encryption (ABE); secure hash algorithm (SHA-1); anonymity; privacy; sensitive parameters


I. INTRODUCTION
Medical domain owes to be attractive region for researchers, challenging them in various aspects like disease prediction, Drug prediction, Drug Repositioning and many more. The recent research focused on disease and treatment prediction using medical repositories accessed and published over distributed environments [1]. This Medical records accessed by several authorities put forth's queries questioning the security and privacy of health care data. The cause entailing such need divulged from the fact that health care data is outsourced for various reasons at the risk of compromising privacy requirements like confidentiality [2,3], Integrity, keyword privacy, authentication audit ability and further. Additionally evolving technologies of information and communication attracts medical domain for integrating health data with technology [4] from domain's like Hospitals, health insurance firms and research laboratories leading to e-Health. E-Health, an attractive domain in recent times that overlaps public health and medical informatics with corporate sectors over internet aims at improving data analysis of health care data locally and worldwide. The cloud successfully offers few advantages over network primarily in enhancing patient care by supporting interaction [5] with healthcare authorities and availability of patient data for analysis and diagnosis [6]. e-Health also offers support for medical research in disease treatment prediction with extended monitoring of epidemics. Further, it helps in cost reduction for engaging expensive hardware, software and data storage at premises.
Eventually e-Health showcases few pullbacks like complexity in interoperability i.e. lacking standard for synchronization, security and privacy issues [20] in shared and public environments [8], regulation controversies related to social and valid frameworks and reliability considerations. They even need to work hands with sensor networks [9] involving data collection. The Healthcare providers in practical are surrounded with several risks [10] from digital technology on cloud despite they encircled advantages.
Among aforementioned issues security and privacy challenges demand utmost attention for realization of its effective utilization. Data confidentiality, authentication and Integrity are at risk in distributed environment. The goal of medical data shared and stored over internet is to provide consistency and high level of security. Despite numerous cryptographic and non-cryptographic methodologies available for enduring security and privacy of e-health data [11,12], few unturned grains are hindering hurdles constraining performance. Our approach proposes a way forward for contending the security and privacy gaps in e-health [13].
The proposed architectural framework attains the goal of security using a Hybrid ABE as well as provides selective access to records based on user predefined access policies like authorized uses, restricted users and un authorized users. The Hybrid ABE provides efficient performance using secure hash algorithm (SHA-1) and Anonymity approaches. The hybrid approach promises high degree of availability, reliability, and efficiency in protecting patient sensitive information [14] upon implementation. Explicitly the architecture gives room for desired authentication to medical archives with extended control for medical stakeholders in general and emergency scenarios on demand. Database Forensic Investigation and proposed harmonized DBFI process using systematic approach with higher certainty. SupriyaMenon M and Rajarajeswari P, 2018, reviewed privacy issues of personalized and context aware privacy and proposed a model for context aware privacy. Jitendra Kumar and Ashutosh Kumar Singh, 2017, came up with a workload prediction model using Long short term memory (LSTM) and tested over three web log datasets proving enhanced accuracy by proposed approach. Supriyamenon M and Rajeswari P, 2020, addressed the complications related to drug repositioning and came up with a hybrid ACO approach enhancing Drug consumption similarities for better repositioning addressing the need for secure patient data. Ma, H., Zhang, R., & Yuan, W, 2016, contributed a mode;l for ABE based Anonymity for Identity revelation.

A. ABE
ABE is an encryption scheme, where the generated cipher text is an outcome relying on user private key and attributes of user data. This public key encryption technique renders plaintext at requested site with decryption supported upon attribute matches of user key and cipher text attributes [15] from attributes of metadata. Although initially introduced in its basic form, exploring amendments of attribute based encryption [16] with multiple authorities involving in user private key generation are also available. ABE has its wide spread usage in several areas like vector driven search engine interfaces, log encryption avoiding log encryption with all recipient keys.
There are two forms of ABE one for key policy KP-ABE and other for cipher policy CP-ABE. The KP-ABE generates user private depending on access tree related to user privileges and encrypting over a set of attributes using algorithms like AES [17]. However cipher text based ABE encrypts user data and attribute with secret keys generated from access trees. ABE rising to be a well preferred mechanism is surrounded with overwhelming challenges like in efficient attribute revocation mechanism, improper key co-ordination, key escrow deficiency and issues related to key revocation mainly for healthcare systems [18] .Few extended problems in the path of ABE is its centralized concept. The need for a centralized body or authority participating in private key generation, makes ABE encounter the flaws due to lack of decentralization. These risks bring down the performance of ABE. One more factor of concern affecting the ABE is speed, which downtrends compared to others due to delay of policy tree construction and computational delay at decryption site also adds upon the issue.

B. SHA-1
The secure hash algorithms enable the determination of Message Integrity that facilitate creation and validation of digital signatures. Digital signatures provide secure security service of Authentication [19] hereby avoiding Denial attacks and repudiations both at source and destination. SHA-1 belongs to the family of secure hash algorithms that generate a hash value known as message digest to facilitate security [21]. It promises its wide spread excellence in several security 158 | P a g e www.ijacsa.thesai.org (IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 12, No. 4, 2021 protocols, mail protocols, TTL, SSL, IPSec and many more. The basic version of the algorithm produces a 160 bit message digest which well prevails against Brute force attack. This variant is considered to be the fastest one but more prone to collision problem, those were overruled in the successor variants. Few well known variants of secure hash algorithms are SHA -2 and SHA-3. The former uses a set of 6 hash functions with digests of size 224, 256, 384, and 512 bits. Among the aforementioned digests SHA-256 and SHA 512 exhibit uniqueness in the sense of computing with 32 bit and 64 bits respectively. They project the variation in the basic shift and additive operations performed. SHA-2 being advanced faced a strong battle to take over its Predecessor. The later addressed SHA-3 by NIST provides compatibility with the former.

C. Anonymization
Anonymization is a process that aims at encapsulating identifying information in a way intending privacy protection. Hence the original data remains anonymous enabling data sharing and transmission among agencies reducing risk of unwanted disclosures [22]. Despite such secure transformations anonymous data never promise to anonymous over time. Several approaches and clever techniques exist that disclose data leading to be de-anonymizated. To handle all such loop holes, several forms of Anonymity are available like k-Anonymity, I-Anonymity, t-closeness, p-sensitivity and many more variations. In k-anonymity, anonymization is a key feature using certain cryptographic hashing. K-Anonymity further has its extension to an (α, k)-anonymity model for privacy preserving data publishing, where α being a fraction and k an integer. The frequency of sensitive value is no more than α. It aims at data security and privacy with further extension to Human and Societal aspects of security and privacy.

IV. PROPOSED APPROACH
Huge amounts of data filling the health care repository is triggering several challenges in due response to providing services. These services claim that cloud computing techniques provide everything as a service i.e. storage as well as security as a service. The major issue of concern is medical confidentiality, portraying the healthy relationship of trust among patients and doctors. The medical data stored in cloud is at high risk of being vulnerable to attacks with irretrievable loss to users with their sensitive data dumped at entrusted servers. With an intension of addressing the above mentioned issues related to data privacy we propose an hybrid approach that resolves the complications in data transmission and provide security.
Phases in proposed approach are discussed below.

Phase 1:
This phase of the proposed system initiates with generation of metadata for the patient records. The attributes in patient records are analyzed and access control structure is defined considering different threshold parameters for various groups of users using ABE approach. Certified attributes defined in the access policy determines which block of plain text should be decrypted for the users with predefined threshold credentials. Elicited from the defined access policies, users are assigned access permissions to the available records.

Phase 2:
The medical records blocks are encrypted considering four randomized algorithms in ABE as Setup, Key generation, Encryption and Decryption.
Setup: At initialization the system generates 2 groups GR1 and GR2 based on security parameters with p prime value, t threshold and b bilinear pairs.
The centralized authority generates master key M K and public key P K by randomly selecting x, u1, u2,…u n Z q where q is the prime number and Z q multiplicative modulo.
Key generation: The authorized authority generates secret key S k for users by using SHA1 with modified feistel structure where SHA1 converts attributes into matrices considering m rows and n columns, where m is the number of attributes and n is the size of SHA output.
The algorithm for key generation is presented below. • Bifurcating Ri mxn to equal partition matrices R ia and R ib .
• Apply transposition resulting in R ia nxm and R ib nxm and add them to Q mxn .
• RO mxn a resulting transpose of Q mxn .
• Revised Le mxn is RO mxn and Ri mxn is previous value of Le mxn until n holds old value.

Process terminates.
Encryption: Sender encrypts the message with key extracted from attributes. C i = Encryption ( P k , PT, A) where P k public key from attributes A for plaintext PT.
Decryption: The Receiver decrypts Ciper C i using the Secret key S k generated by SHA1.
ABE explicitly supports threshold operations on attributes to specify permitted access control structures to the users of different groups.

Phase 3:
Among different groups of users with certain attribute combinations, the limited access groups of users considered as restricted users are subjected to feasible Anonymity technique 159 | P a g e www.ijacsa.thesai.org (IJACSA) International Journal of Advanced Computer Science and Applications, Vol. 12, No. 4, 2021 with lower distortion. The k-anonymity technique preferred avoids identity disclosure.
Algorithm for ( , )Anonymity : Input: Raw table   Output: Hybrid Anonymity table. 1. Initialization stage Generate user for input vector and for array of users considered 1,2,3,…n, compute dissimilarity matrix DMT by calculating distances.
2. Anonymization at Client side 1. Compute DMT. 2. Assign false to all points and select a point p with C c as centroid of user c i and mark it as true.
3. Consider false points as minimum distance from C,with social attributes S A . 4. Add above considered point to P, and check the frequency with Anonymization parameter k.
If frequency of (S A ) < k { Consider and adjust centroid; } Else Abort; 4. Repeat until all points in c are verified and return. 5. Group unassigned points to nearest user and ensure user satisfying (1,K1) anonymity.
3. Anonymization at Server side 1. Consider the nearest user pairs p1 and p2 in client side matrix. 2. Combine p1and p2. 3. Size of u=p1+p2. 4. Compute representative vector R* u using tree access structure T. 5. T y denotes sub-tree where y is root in tree T.
When root node is r in T= T r Attribute set =e that confirms to access policy of T y Then T y (e) =1 If y is a leaf node and attribute attr (y) T y (e) =1 Else Validate sub-nodes. 6. Repeat process until each user satisfies (a2, k2) anonymity.
Based on the Anonymity levels attained from proposed technique predefined threshold attributes in access tree structure are sent in Plain text and other blocks are anonym zed.

V. PERFORMANCE EVALUATION AND ANALYSIS
Our Proposed Hybrid ABE Approach projects efficient performance with respect to time and Memory Utilization when compared with existing techniques like Common Database Forensic Investigation Process (CDFIP), Real-time Operational Data Base (RODB) Extraction-Transformation-Loading (ETL), and Long Short Term Memory (LSTM) [7] to Recurrent Neural Network (RNN). The Simulation parameters considered for Implementation of the proposed approach are shown in the Table I. To evaluate the performance of Hybrid ABE, Encryption time, Decryption time and Memory Utilization are considered.
Encryption time: This computes the throughput of the encryption scheme with respect to user instances and encryption time.
Decryption time: This computes the throughput of the decryption scheme with respect to user instances and decryption time.
Memory Utilization: This evaluation parameter projects the average utilization of system memory in bytes for different user instances. Table II shows an improved performance of time for different user instances using Hybrid Approach against existing methods opted.           Hence the simulation results of the proposed algorithm outperform in terms of Performance, Encryption and Decryption throughput and memory utilization providing improved Privacy for patient sensitive data.

VI. CONCLUSION
This paper aimed to discuss the importance of security of patient data based on the access priorities of users, using a Hybrid ABE Approach. In due course several techniques related to mobile healthcare and e-healthcare grabbed concentration in research, but lacked profound architecture to preserve patient data. Our framework offers a innovative and 161 | P a g e www.ijacsa.thesai.org qualitative technique using SHA-1 and improved Feistel network in key generation ensuring authentication, and confidentiality during transmission entailing limited access to user communities considering access policy. The groups of users with limited access are subjected to Anonymity techniques. The result of our method renders improvised performance in several evaluation parameters considered. Lastly, we conclude that the roadmap presented endeavors a feasible solution for discussed privacy issues.