Data in the time of COVID-19: a general methodology to select and secure a NoSQL DBMS for medical data

NoSQL database management systems classification

NoSQL database management systems are classified into five categories, namely Document-based Stores, key-value stores, Column-based or wide column Stores, Graph-based Stores, and Object Stores. Multi-model database categories may be added as shown in Fig. 1.

NoSQL DBMSs comparison criteria

Database management systems will be analyzed according to their basic information, security mechanisms, and potential attacks.

• Basic information includes:

  • The owner and year of first appearance

  • The classification, specifying whether it is open source or proprietary

  • The DBMSs implementation programming language

  • Some characteristics of the database management system

  • The query language used in each database management system.

• Security mechanisms

  • Authentication is the process of verifying the entity identity (user or device) permission to use the resources (data, application, etc.) (Zahid, Masood & Shibli, 2014).

  • Verification is usually achieved through accessing control mechanisms, such as users’ passwords and certificates (Grolinger et al., 2013).

  • Authorization is the mechanism that ensures that only an authorized user is allowed to access system resources. Authorization is usually performed with the user’s permission (Grolinger et al., 2013).

  • Auditing is the process of keeping track and recording all actions performed by database users (Elmasri & Navathe, 2017).

  • Data Encryption refers to protecting and encrypting sensitive data throughout the communications network so it cannot be read by unauthorized users or attackers (Elmasri & Navathe, 2017; Grolinger et al., 2013). There are three levels of encryption solution; Data at rest, Client-to-server communication, and Server-to-server connections (Grolinger et al., 2013).

• Attacks: some attack types that DBMSs may endure include:

  • Script Injection, wherein a security vulnerability is used by an adversary to inject malicious code in pages or web forms to damage or retrieve data (Sullivan & Vincent, 2011; Jim, Nikhil & Michael, 2007).

  • SQL Injection, which may occur when an adversary injects a string input through the application to damage, change the database, retrieve the sensitive data, and run commands to deny service (Elmasri & Navathe, 2017).

  • Denial of service (DOS), in which an adversary floods the server with requests, thus preventing legitimate users from accessing the service, or the attacker may delete some data (Elmasri & Navathe, 2017).

  • An inactive connection may occur when an adversary opens the connection for a long time.

MongoDB

• Introduction:

  MongoDB is an open-source document-oriented NoSQL DBMS. It is a schema-free, highly available, fault-tolerant and scalable NoSQL DBMS. MongoDB supports sharding by configuring shared clusters (Zahid, Masood & Shibli, 2014). It was developed by MongoDB Inc., first appearing in 2009; its stable release appeared in June 2018. MongoDB was written in C++, C, and JavaScript (Nayak, Poriya & Poojary, 2013; Özsu & Valduriez, 2020), and it consists of one or more collections of documents. A collection is analogous to a table, but has no pre-defined schema. The document, which is the data storing unit in MongoDB, has an ID and is equivalent to a record in the relational DB. Insert, delete, and update operations can be performed on a collection (Dindoliwala & Morena, 2017). Mongo Query Language is the query language used By MongoDB DBMS to manipulate certain documents from a database collection. Map-reduce and REST querying are supported by MongoDB (Chodorow, 2013).

• Security Analysis:

• Authentication: granting to users through the database itself or integration with an external mechanism like LDAP (Lightweight Directory Access Protocol). MongoDB grant Authentication to entities by using the SCRAM-SHA-1 IETF standard (Zugaj & Beichler, 2019). An authentication command is used to authenticate the users, whereas database-nodes are authenticated to the MongoDB cluster via key files. The default Authentication mechanism in MongoDB is SCRAM-SHA-1, which verifies the supplied user credentials with the name of the user, user’s password, and Authentication database. With LDAP, MongoDB can authenticate and authorize users directly via corporate LDAP infrastructure to enforce centralized access policies (Dindoliwala & Morena, 2017). MongoDB doesn’t support authentication, however, if running in shared mode (Sahafizadeh & Nematbakhsh, 2015; Noiumkar & Tawatchai, 2014). When the authentication in MongoDB is based on the key or the password, it’s called a pre-shared secret. This password has been hashed with the MD5 algorithm before being stored in the key file. Hackers can crack the pre-shared secret’s hashed value by cracking the MD5 (Dindoliwala & Morena, 2017), because the MD5 algorithm is not secured (Sahafizadeh & Nematbakhsh, 2015). MongoDB uses SSL with X.509 certificates to secure communication between the user and the MongoDB cluster and intra-cluster authentication (Zahid, Masood & Shibli, 2014).

• Authorization: Administrators in MongoDB can determine the permissions for users or applications. They can also decide what data can be seen by users or applications when performing a database query. The authorization mechanism in MongoDB contains LDAP authorization, roles, and Field-Level Security with Read-Only Views. In MongoDB privileges are assigned to roles, then roles are assigned to users. MongoDB involves built-in roles that are used by administrators to control access to the MongoDB system. A privilege consists of a specified resource and actions permitted on this resource. A resource is a collection, set of collections, databases, or a cluster. MongoDB supports a simple role-based authentication system that allows administrators to decide who has access to the database and which level of access they have. Furthermore, MongoDB supports authorization through LDAP (Dindoliwala & Morena, 2017).

• Auditing: Auditing framework stores actions such as operations that have been performed by the users in the database, and authentication and authorization activities along with write and read operations on the database. Administrators can extract and filter audit trails for any operation with MongoDB. For example, it is possible to audit and log the identities of users who accessed specific documents and the changes they made to the database during their session. Administrators can configure MongoDB to log all the actions or specific actions. MongoDB Enterprise Advanced (Dindoliwala & Morena, 2017) also supports the role-based auditing. With MongoDB, it is possible to report the DB activities as well.

• Data Encryption: in MongoDB, data can be encrypted over the network, in backups, and the rest in permanent storage. MongoDB Enterprise Advanced supports FIPS 140-2 encryption when it run in FIPS Mode (Dindoliwala & Morena, 2017). In MongoDB, the data files are not encrypted automatically, so all data is stored as plaintext. This means that any malicious user with access to the file system can extract information from the files. MongoDB uses RESTful to manage its server via the HTTP protocol. However, there is no data encryption for these ports, which means that hackers can display, capture and get access to the data since no data encryption was performed while the data was being sent back and forth between the client and server (Noiumkar & Tawatchai, 2014).

• Script Injection: Using JavaScript, hackers can attack MongoDB. Examples about the attacks that occurred on MongoDB can be found in this reference (Jim, Nikhil & Michael, 2007). NoSQL databases are exposed to injection attacks like SQL databases. Son presented four examples of GitHub site for injection vulnerabilities in PHP MongoDB-based applications (Son, McKinley & Shmatikov, 2013).

CouchDB

• Introduction:

  CouchDB is an open-source NoSQL DBMS. It was developed using Erlang (George, 2013). It became an Apache project in 2008. It was released in 2005 and the stable release appeared in November 2017. CouchDB runs on Hadoop Distributed File Systems (HDFS). CouchDB is document-based NoSQL DBMS, fault-tolerant, highly scalable, available, and flexible (Sahafizadeh & Nematbakhsh, 2015). It supports document redistribution across nodes by cluster configuration for large performance improvements. A cluster uses the phenomenon of incremental replication by periodically copying any changes in a document on a single node to other nodes. The incremental replication results in inconsistency and data redundancy in the cluster. CouchDB uses the JavaScript, JSON, and SQL++ as its query language (Özsu & Valduriez, 2020). Map-reduce and REST querying are supported by CouchDB as well (Noiumkar & Tawatchai, 2014).

• Security Analysis:

• Authentication: CouchDB uses a process called CRUD to perform authentication (Noiumkar & Tawatchai, 2014). The authentication in CouchDB is based on both cookies and passwords. It uses the PBKDF2 hash algorithm to encrypt the password and send it over the network using SSL protocol (Sahafizadeh & Nematbakhsh, 2015; Zahid, Masood & Shibli, 2014).

• Authorization: There are three levels of users within the CouchDB; database admin, server admin, and database member (Grolinger et al., 2013). Authorization is implemented only at the database level where only a single role in access control is supported (Zahid, Masood & Shibli, 2014).

• Auditing: The level of auditing in CouchDB is moderate (logging of all changes to user-profiles and the level of database). Auditing is provided to log events and views in log files. However, CouchDB does not support automatic backups of database logs, replicas, and automatic logging; the configuration of logs is thus the responsibility of the database administrators (Zahid, Masood & Shibli, 2014).

• Data Encryption: CouchDB does not have automatic data encryption, so the data files are at risk of being accessed and read directly (Sahafizadeh & Nematbakhsh, 2015; Noiumkar & Tawatchai, 2014). Data Encryption is supported in Client/Server via SSL. Also, data encryption could be supported using HTTPS connections in Server/Server (Grolinger et al., 2013).

• Denial of service attack: Websites such as securityfocus.com display vulnerabilities that exist only on Apache CouchDB 1.5.0 (SecurityFocus, 2020). These vulnerabilities are used by hackers to crash it. So CouchDB is potentially vulnerable to Denial of service attacks (Sahafizadeh & Nematbakhsh, 2015).

• Script Injection: CouchDB uses JSON to manipulate the data, making it vulnerable to Script Injection. In other words, hackers can launch JSON Injection to attack CouchDB (Sahafizadeh & Nematbakhsh, 2015; Noiumkar & Tawatchai, 2014).

Couchbase server

• Introduction:

  Couchbase Server is an open-source, document-oriented NoSQL DBMS. Couchbase Server was developed by Couchbase Inc., initially released in August 2010 with the stable release launched in February 2018. Couchbase Server was developed using Erlang, C, and C++ (George, 2013). Various data centers contain all the cluster servers. The documents are stored in vbuckets, which are special data containers uniformly distributed across the cluster. More nodes can be added and removed because Couchbase cluster scales are completely horizontal (Zahid, Masood & Shibli, 2014). Map-reduce and REST querying is supported by the Couchbase server (Grolinger et al., 2013). Couchbase also supports SQL++, an SQL like language which has been designed to become a unified query language for NoSQL system.

• Security analysis:

• Authentication: SASL (Simple Authentication and Security Layer) is supported by name and password on Couchbase Server HTTP. Basic authentication is used in Couchbase management REST (Representational State Transfer) API (Application Programming Interface) (Grolinger et al., 2013). External Authentication is also supported (Zahid, Masood & Shibli, 2014).

• Authorization: Couchbase Server does not support authorization (Grolinger et al., 2013).

• Auditing: Couchbase Server does not support auditing (Grolinger et al., 2013).

• Data Encryption: Couchbase Server does not support data Encryption in Data at rest, Client/Server, and Server/Server (Grolinger et al., 2013).

Redis

• Introduction:

  Redis is an Open source NoSQL DBMS (Grolinger et al., 2013), developed using C and C++ (Meier & Kaufmann, 2019) by Salvatore Sanfilippo. Redis was initially released in May 2009 and the stable release appeared in June 2018. It is a Key-value memory database, So, Redis data will be loaded into memory (Dayan, Manos & Stratos, 2018) when it runs and all operations are run in memory. Redis periodically saves data asynchronously to the hard disk. It achieves high-performance thanks to its use of pure memory. Redis can handle more than 100,000 write or read operations per second. The maximum limit of value is 1GB. However, Redis cannot be used as big data storage, and scalability is poor because the capacity of the Database is limited by the physical memory. As such, Redis is suitable for providing high-performance computing for a small amount of data (Han et al., 2011). SQL-like and Map-reduce querying is not supported in Redis but REST querying is supported by third-party APIs (Grolinger et al., 2013).

• Security analysis:

  • Authentication: Redis provides a password-based authentication to its clients. Passwords are set by system administrators and stored in plaintext format. Redis does not provide default authentication and listens on all IP addresses on port 6739 (Han et al., 2011).

  • Authorization: Redis does not support any kind of access control (Han et al., 2011) i.e., it does not support authorization (Paterson et al., 2006).

  • Auditing: Redis does not provide support for auditing (Tudorica & Bucur, 2011; Paterson et al., 2006).

  • Data Encryption: Redis does not provide support for Data encryption. It stores data in plaintext (Paterson et al., 2006). As such, hackers—or anyone who can access the Redis server - will be able to read all the data in the database. No data encryption is performed in the communication between client and server, and between other servers, neither at the same cluster nor different clusters (Nayak, Poriya & Poojary, 2013). Redis does not have data-at-rest encryption (Zugaj & Beichler, 2019).

  • Denial of service attack: Denial of service attacks on Redis was not reported (Nayak, Poriya & Poojary, 2013).

  • Script Injection: Redis does not have a concept of string escaping, so injection becomes impossible (Nayak, Poriya & Poojary, 2013; Paterson et al., 2006).

Voldemort

• Introduction:

  Voldemort is an Open source key-value NoSQL DBMS, developed by LinkedIn (Dayan, Manos & Stratos, 2018). Its initial release was in 2009 and the stable release appeared in July 2017 (Grolinger et al., 2013). Voldemort was developed using Java (Deka, 2013). Three simple operations are included in Voldemort; reading, writing, and deletion. All of them are executed using a key (Chen, Shiwen & Yunhao, 2014). There is no SQL-like querying language in Voldemort but Map-reduce querying is supported (Grolinger et al., 2013).

• Security Analysis:

  • Authentication: There is no authentication mechanism in Voldemort.

  • Authorization: There is no authorization mechanism in Voldemort.

  • Auditing: There is no auditing in Voldemort.

  • Data Encryption: No encryption, neither between the client and the server nor among the servers (Grolinger et al., 2013).

Oracle Berkeley DB (BDB)

• Introduction:

  Oracle Berkeley DB is a closed source key-value NoSQL DBMS. Oracle Berkeley DB was developed by Sleepycat Software. Oracle Berkeley DB initial release appeared in 1994 and the stable release was in June 2018. It is a high-performance data management services to applications. It uses a simple function for data access and management (Sullivan & Vincent, 2011). SQLite and REST querying is supported by Oracle Berkeley DB, but Map-reduce querying is not (Grolinger et al., 2013). In terms of implementation of Oracle Berkeley DB, was implemented in C, Oracle Berkeley XML was implemented in C++ and Oracle Berkeley JE was implemented in Java.

• Security Analysis:

  • Authentication: There is no authentication mechanism in Berkeley DB.

  • Authorization: There is no authorization mechanism in Berkeley DB.

  • Auditing: There is no auditing in Berkeley DB.

  • Data Encryption: Berkeley DB supports data at rest encryption. No encryption between Servers (Grolinger et al., 2013).

Neo4j

• Introduction:

  Neo4j is an open-source graph NoSQL DBMS (Paterson et al., 2006), developed by Neo Technology. It was initially released in 2007 and the stable release appeared in April 2018. It was developed using Java. The advantages of Neo4j include providing object-oriented, flexible network structure, reliability, highly available, and scalability. Neo4j uses the graph data model which consists of nodes and edges and the relationships between them. Neo4j uses CYPHER query language. REST querying is supported by Neo4j but Map-reduce is not supported (Grolinger et al., 2013). Neo4j is used extensively in software that has complex relationships such as social networking and recommendation engines (Moniruzzaman & Hossain, 2013).

• Security Analysis:

  • Authentication: Neo4j does not support Authentication (Grolinger et al., 2013).

  • Authorization: Neo4j does not support authorization (Grolinger et al., 2013; Paterson et al., 2006)

  • Auditing: Neo4j does not support auditing (Grolinger et al., 2013; Paterson et al., 2006).

  • Data Encryption: Neo4j does not support data at rest encryption, nor does it support data encryption between servers (Grolinger et al., 2013); however it uses SSL-based communication protocol between the client and the server (Paterson et al., 2006).

HyperGraphDB

• Introduction:

HyperGraphDB is an Open source NoSQL DBMS that implements powerful knowledge management formalism known as directed hypergraphs (Grolinger et al., 2013). HyperGraphDB is developed by Kobrix Software, Inc. The stable release of HyperGraphDB appeared in May 2017. Hyper-GraphDB was developed using Java. Higher-order Relations are naturally represented by the HyperGraphDB model. The HyperGraphDB model is very useful in areas related to data modeling such as artificial intelligence and knowledge of representation and bioinformatics. This data model is designed for AI, Knowledge representation and the semantic Web (Kobrix Software Inc, 2020; Angles, 2012). HyperGraphDB provides SQL- like query language. REST querying is supported by HyperGraphDB but Map-reduce is not (Grolinger et al., 2013).

• Security Analysis:

  –Authentication: HyperGraphDB does not support authentication.

  –Authorization: HyperGraphDB does not support authorization.

  –Auditing: HyperGraphDB does not support auditing.

  • Data Encryption: HyperGraphDB does not support data encryption (Grolinger et al., 2013).

AllegroGraph

• Introduction:

  AllegroGraph is a closed-source graph NoSQL DBMS, developed by Franz Inc. The stable release appeared in October 2017 (Khasawneh, AL-Sahlee & Safia, 2020; Grolinger et al., 2013). It was developed using Common Lisp (Franz Inc, 2020). It offers high-performance and is mainly used for developing semantic web applications. It can store data and meta-data as RDF (Resource Description Framework) triples. AllegroGraph maintains high performance when accessing billions of quads due to efficient memory utilization along with disk-based storage (Arora & Rinkle, 2013; Franz Inc, 2020). SPARQL and Prolog are the query languages used by AllegroGraph. REST querying is supported by AllegroGraph but Map-reduce is not (Grolinger et al., 2013).

• Security Analysis:

• Authentication: Authentication is supported by AllegroGraph.

• Authorization: Authorization is supported by AllegroGraph such as read, write, and delete.

• Auditing: AllegroGraph can be used to record specific changes in the audit log.

• Data Encryption: AllegroGraph does not support data at rest encryption. But it uses an HTTPS-based communication protocol between the client and the server (Grolinger et al., 2013).

Hypertable

• Introduction:

  Hypertable is a column-oriented NoSQL DBMS. It was developed by Zvents before 2008 and the stable release appeared in March 2016. Hypertable was developed using C++ (Deka, 2013). It is a high-performance open-source DBMS, and can run on Hadoop Distributed File Systems (HDFS). It is modeled after Google’s Bigtable DBMS. Like Bigtable, Hypertable stores data using the column-oriented methodology (Khasawneh, AL-Sahlee & Safia, 2020). There is no SQL-like query language in Hypertable.

• Security Analysis:

• Authentication: Hypertable does not support Authentication (Nayak, Poriya & Poojary, 2013; Paterson et al., 2006). Moreover, it does not support Authentication in communication between the client and the server or among its servers (Nayak, Poriya & Poojary, 2013).

• Data Encryption: Hypertable does not support encryption for its data files (Nayak, Poriya & Poojary, 2013; Paterson et al., 2006). Hypertable does not support encryption in communication between the client and server or between its servers (Nayak, Poriya & Poojary, 2013).

• Denial of service attack: There is no information reported about denial of service attacks on Hypertable.

• Script Injection: Though Hypertable has an HQL (Hypertable Query Language), which is similar to the SQL, it is not vulnerable to script injection (Nayak, Poriya & Poojary, 2013).

HBase

• Introduction:

  HBase is an open-source column-oriented NoSQL DBMS. HBase was developed by the Apache Software Foundation (2020). Its initial release was in February 2015 and the stable release was in April 2018. It was developed using Java (Meier & Kaufmann, 2019). Features include automatic distributability and scalability. HBase is based on the concept of Google’s Bigtable and implemented in Java (Tudorica & Bucur, 2011; Paterson et al., 2006). There is no SQL-like query language in HBase, but Map reduce and REST querying are supported (Grolinger et al., 2013).

• Security Analysis:

• Authentication: HBase supports token-based Authentication, for MapReduce tasks and user Authentication is done by using SASL (Simple Authentication and Security Layer with Kerberos Tudorica & Bucur, 2011; Paterson et al., 2006).

• Authorization: HBase supports authorization by access control list. Permissions include create, read, write and admin (Grolinger et al., 2013; Paterson et al., 2006).

• Auditing: Auditing is supported by HBase (Han et al., 2011).

• Data Encryption: HBase does not support the encryption of data at rest (Tudorica & Bucur, 2011; Paterson et al., 2006), but HBase supports the encryption in communication between client and server (Grolinger et al., 2013).

• Denial of service attack: No report for denial of service attack.

• Script Injection: No report for script injection (Paterson et al., 2006).

Gemstone

• Introduction:

  Gemstone is an open-source an Object NoSQL DBMS. It was developed by GemTalk Systems. Its first appearance was in 1986 and the stable release is 64 Bit 3.3.7. Gemstone is Proprietary commercial software. Multiuser environments are supported by Gemstone; each user can have one or more sessions, and multiple user sessions can be active at the same time. Gemstone provides security at several levels from login authorization to object access (Dindoliwala & Morena, 2017). There is no general query language like SQL in Gemstone.

• Security Analysis:

• Authentication: In Gemstone, each user is identified by a unique user ID and a password. The user is represented by an instance of the class User Profile. The User Profile contains a user ID, password, default authorization information, and the user’s group. Only users who have a User Profile can logon to the system.

• Authorization: Authorization exists within Gemstone and controls individual object access.

• Auditing: The DB administrator can also configure the Gemstone system to monitor failures to login.

• Data Encryption: Gemstone does not support Data Encryption (Dindoliwala & Morena, 2017).

db4o

• Introduction:

  db4o is an open-source object NoSQL DBMS. It was developed by Action and the stable release 8.0. db4o was developed using Java and C# (Noiumkar & Tawatchai, 2014). db4o-SQL is an interface to allow SQL queries into a db4o database.

• Security Analysis:

• Authentication: db4o provides only an internal authentication mechanism.

• Authorization: db4o offers file-level authorization.

• Auditing: db4o lacks auditing capabilities.

• Data Encryption: db4o uses the eXtended Tiny Encryption Algorithm (Dindoliwala & Morena, 2017) for data encryption.

Objectivity/DB

• Introduction:

  Objectivity/DB is an object NoSQL DBMS. Objectivity/DB is proprietary to Objectivity Inc. and first commercialized in 1990. Objectivity/DB was developed using Java, C#, C++, and Python. Since the Objectivity/DB schema is created from internal class definitions from the programming language, it is easier to maintain changes to the schema (Dindoliwala & Morena, 2017). Objectivity/DB provides OQL (Object Query Language which is like SQL querying language

• Security Analysis:

  • Authentication: For Authentication Objectivity/DB has two mechanisms; the first (by default) is Kerberos; the second is by using Advanced Multithreaded Server (AMSprotocol).

  • Authorization: Relies on the operating system and file systems for access control.

  • Auditing: No auditing capabilities have been mentioned in Objectivity/DB.

  • Data Encryption: Objectivity/DB doesn’t provide any inbuilt encryption (Dindoliwala & Morena, 2017).

Cassandra

• Introduction:

  Cassandra is an open-source Multi-model database NoSQL DBMS. It was developed by Apache Software Foundation and its stable release appeared in February 2018. Cassandra was developed using Java, and is used by Facebook. Cassandra is a Multi-model database of both column stores and key-value stores (Moniruzzaman & Hossain, 2013). Its key characteristics include: a free schema, flexibility, convenience of adding or deleting fields, support for range queries, partition tolerance, and high scalability (Han et al., 2011). Cassandra has three basic components; data, centers, and clusters nodes. The data in a cluster is organized into key spaces (databases), which contain tables. Tables contain rows, and rows have columns. The query language used is Cassandra Query Language (CQL) also Cassandra supports Map-reduce querying and REST querying by Third-party APIs (Dindoliwala & Morena, 2017).

• Security Analysis:

  • Authentication: authentication in Cassandra is particularly weak since passwords are encrypted using the MD5 hash. The entire authentication in Cassandra is provided between the client and the Cassandra cluster. In other words, the inter-node message exchange does not support Authentication by default. Hence, a malicious user with access to the network used by the Cassandra cluster can bypass the client-side Authentication and extract data or damage it. However, the transmission security at the data center, rack, and a cluster is provided by Cassandra via enabling SSL/TLS in its configurations (Han et al., 2011). Cassandra supports only internal Authentication mechanisms. Cassandra 3.0 Authentication is role-based and stored internally in Cassandra system tables (Dindoliwala & Morena, 2017).

  • Authorization: The entire authorization mechanism in Cassandra is provided between the client and the Cassandra cluster (Han et al., 2011). As such, a malicious user with access to the network used by the Cassandra cluster can bypass the client-side authorization and extract data or damage it. In 2015, it had been proven that Cassandra did not support authorization (Sahafizadeh & Nematbakhsh, 2015; Paterson et al., 2006). Cassandra uses the GRANT/REVOKE mechanism to manage permissions as part of the authorization mechanism. In Cassandra 2.2 and later versions, Role-based access control is available and permissions may be applied on resources such as keyspace, table, function (Dindoliwala & Morena, 2017).

  • Auditing: Cassandra does not support auditing (Tudorica & Bucur, 2011; Paterson et al., 2006). Cassandra auditing is available in Enterprise Cassandra as a log4jbased integration and a per-node basis. Filters are available for logging using a combination of the following categories—ADMIN, AUTH, ALL, DCL, DML, and QUERY (Dindoliwala & Morena, 2017).

  • Data Encryption: Data files in Cassandra are stored without encryption and the database does not have automatic data encryption, so malicious users can access the data and read it (Nayak, Poriya & Poojary, 2013; Tudorica & Bucur, 2011). Since version 3.2, Cassandra supports at-rest data encryption through Transparent Data Encryption (TDE) (Dindoliwala & Morena, 2017).

  • SQL Injection: Cassandra Query Language (CQL) has a similar syntax to SQL’s, so it is believed that it can be attacked in the same way SQL is vulnerable to attacks by SQL injection (Nayak, Poriya & Poojary, 2013; Paterson et al., 2006).

  • Denial of service attack: Cassandra performs one thread per client. As such, it is vulnerable to denial of service attacks. A malicious user creating enough fake connections can drain Cassandra’s resources. If hackers knew the IP addresses of all Cassandra servers in the cluster and created enough fake connections, they could drive it out of service (Nayak, Poriya & Poojary, 2013).

  • Inactive connections: Cassandra does not set the timeout value for inactive connections; consequently, it has a problem in managing inactive connections (Nayak, Poriya & Poojary, 2013; Paterson et al., 2006).

DynamoDB

• Introduction:

  DynamoDB is a Closed source NoSQL DBMS used by Amazon. Its initial release appeared in January 2012 (Grolinger et al., 2013). DynamoDB was developed using Java (Meier & Kaufmann, 2019). It is fast and flexible and supports both the document and key-value data models (Paterson et al., 2006). The query language is proprietary in DynamoDB but MapReduce and REST query are supported (Grolinger et al., 2013).

• Security Analysis:

  • Authentication: Authentication is supported by DynamoDB (Paterson et al., 2006). It is ensured by the integration of Identity and Access Management (IAM services Grolinger et al., 2013; Paterson et al., 2006).

  • Authorization: Authorization is supported by DynamoDB (Paterson et al., 2006). It allows users to create policies and operations on domains (Grolinger et al., 2013).

  • Auditing: DynamoDB integrates with Amazon Cloud Watch Service. Access information about latencies for operations, requests throughput, and the amount of data stored (Grolinger et al., 2013) are subject to auditing at the ACWS.

  • Data Encryption: DynamoDB does not support the encryption of data at rest (Grolinger et al., 2013). Communication between the client and the server uses the https protocol (Paterson et al., 2006).