Software-as-a-Service Security Challenges and Best Practices: A Multivocal Literature Review

: Cloud computing (CC) is the delivery of computing services on demand and is charged using a “pay per you use” policy. Of the multiple services offered by CC, SaaS is the most popular and widely adapted service platform and is used by billions of organizations due to its wide range of beneﬁts. However, security is a key challenge and obstacle in cloud adoption and therefore needs to be addressed. Researchers and practitioners (R&P) have discussed various security challenges for SaaS along with possible solutions. However, no research study exists that systematically accumulates and analyzes the security challenges and solutions. To ﬁll this gap and provide the state-of-the-art (SOTA) picture of SaaS security, this study provides a comprehensive multivocal literature review (MVLR), including SaaS security issues/challenges and best practices for mitigating these security issues. We identiﬁed SaaS security issues/challenges and best practices from the formal literature (FL) as well as the grey literature (GL) to evaluate whether R&P is on the same page or if controversies exist. A total of 93 primary studies were identiﬁed, of which 58 are from the FL and 35 belong to the GL. The studies are from the last ten years, from 2010 to 2021. The selected studies were evaluated and analyzed to identify the key security issues faced by SaaS computing and to be aware of the best practices suggested by R&P to improve SaaS security. This MVLR will assist SaaS users to identify the many areas in which additional research and development in SaaS security is required. According to our study ﬁndings, data breaches/leakage, identity and access management, governance and regulatory compliance/SLA compliance, and malicious insiders are the key security challenges with the maximum frequency of occurrence in both FL and GL. On the other hand, R&P agree that up-to-date security controls/standards, the use of strong encryption techniques, regulatory compliance/SLA compliance, and multifactor authentication are the most important solutions.


Introduction
Cloud computing (CC) offers a consolidated pool of configurable computing tools and computing outsourcing processes that enable various computing services to be offered to individuals and organizations. Millions of organizations have adopted CC due to its potential benefits, such as cost efficiency, improved collaboration, scalability, flexibility, automatic software updates, business continuity, etc. [1,2] CC encompasses a wide range of services and implementation models, as shown in Figure 1. The three types of services provided by CC are platform-as-a-service (PaaS), infrastructure-as-a-service (IaaS), and software-as-aservice (SaaS). IaaS provides options such as renting IT storage (virtual or physical) and services provided by CC are platform-as-a-service (PaaS), infrastructure-as-a-service (IaaS), and software-as-a-service (SaaS). IaaS provides options such as renting IT storage (virtual or physical) and networking capabilities, while PaaS provides on-demand product development, training, delivery, and management tools. At the same time, SaaS is a subscription-based way of providing on-demand software applications through the cloud [3,4]. CC has four different models, namely the private cloud (PRC), the public cloud (PBC), the hybrid cloud (HYC), and the community cloud (CMC). In a PBC system, resources are shared by a group of users known as tenants. The cost of using CC is determined by how much IT infrastructure is used. On the other hand, individuals and businesses that want a PRC must have their own dedicated platform that is not shared by others. CMC is a cloud infrastructure used by users in the same industry or with common goals, and the HYC allows data and resources to be exchanged using PBC and PRC settings [5,6]. Security is one of the key challenges in all the platforms provided by CC. However, it has become a significant concern on the public SaaS cloud. The belief that storing sensitive data in a third-party data center leads to various security breaches becomes a primary obstacle to CC adoption. According to the statistics provided by Statista, the PBC SaaS market is increasing rapidly with time (as shown in Figure 2). It is a model in which the CSP hosts applications remotely and makes them accessible to consumers on-demand over the Internet. Customers benefit from the SaaS model in a variety of ways, including increased operating performance and lower costs. SaaS is quickly gaining attention as the preferred distribution model for business IT services. Most businesses, however, are still wary of the SaaS model due to a lack of insight into how their data are processed and protected. Thus, security is the top obstacle to SaaS adoption for enterprise IT infrastructures [7,8]. According to the statistics provided by Statista, the PBC SaaS market is increasing rapidly with time (as shown in Figure 2). It is a model in which the CSP hosts applications remotely and makes them accessible to consumers on-demand over the Internet. Customers benefit from the SaaS model in a variety of ways, including increased operating performance and lower costs. SaaS is quickly gaining attention as the preferred distribution model for business IT services. Most businesses, however, are still wary of the SaaS model due to a lack of insight into how their data are processed and protected. Thus, security is the top obstacle to SaaS adoption for enterprise IT infrastructures [7,8]. Appl. Sci. 2022, 12, x FOR PEER REVIEW 3 of 28 Figure 2. SaaS market size in billion USD as per Statista [9].
Various solutions for CC security in general and SaaS security are provided in the literature. However, no single solution fits all organizations under all circumstances. The security of SaaS is the joint responsibility of the cloud tenant (CT) and the cloud service provider (CSP), but customers still expect 100% security assurance from CSPs [10]. To help the SaaS CT and CSP, there is a need to provide better guidelines. This is only possible if they are aware of all possible security issues and challenges and best practices that may help overcome these challenges. To address this gap, this MVLR aims to provide a detailed overview of the opinions of researchers and practitioners (R&P) regarding SaaS security issues/challenges and best practices for secure SaaS. The objective of this MVLR is to conduct a detailed and systematic review of the scientific literature and the GL to identify SaaS security issues and challenges. This MVLR will try to answer the following research questions: RQ1: What software security challenges are involved in SaaS as identified in the FL?
The aim of this question is to systematically gather and review the formal research studies published in peer-reviewed journals/conferences/workshops that focus on SaaS security issues and challenges and compile a list of potential SaaS cloud challenges. This will help SaaS CSPs and tenants to secure SaaS against internal and external threats. RQ2: What software security challenges face SaaS as identified in the GL?
This research question is designed to know the practitioner's opinions on SaaS security issues/challenges. These challenges will be identified by screening the GL available on the Google search engine. RQ 3: Which practices are suggested by the FL for improving SaaS cloud security?
This research question aims to analyze the selected primary studies to find the best practices that help in improving the security of the SaaS cloud. RQ4: Which practices are suggested by the GL for improving SaaS cloud security?
This question will identify the best practices for improving SaaS security as suggested by practitioners in this area through the analysis of the identified GL. RQ5: Is there any similarity or discrepancy between R&P opinions regarding SaaS security issues and solutions?
This question will compare and analyze the results of the aforementioned four research questions to find the similarities and differences between R&P opinions.
The remaining paper organization is as follows: the background information is covered in Section 2. Section 3 summarizes the related work. The research methodology is defined in Section 4. The study's findings are presented in Section 5, which is accompanied by a review of the findings in Section 6. Section 7 concludes the paper by providing insights into future research. Various solutions for CC security in general and SaaS security are provided in the literature. However, no single solution fits all organizations under all circumstances. The security of SaaS is the joint responsibility of the cloud tenant (CT) and the cloud service provider (CSP), but customers still expect 100% security assurance from CSPs [10]. To help the SaaS CT and CSP, there is a need to provide better guidelines. This is only possible if they are aware of all possible security issues and challenges and best practices that may help overcome these challenges. To address this gap, this MVLR aims to provide a detailed overview of the opinions of researchers and practitioners (R&P) regarding SaaS security issues/challenges and best practices for secure SaaS. The objective of this MVLR is to conduct a detailed and systematic review of the scientific literature and the GL to identify SaaS security issues and challenges. This MVLR will try to answer the following research questions: RQ1: What software security challenges are involved in SaaS as identified in the FL?
The aim of this question is to systematically gather and review the formal research studies published in peer-reviewed journals/conferences/workshops that focus on SaaS security issues and challenges and compile a list of potential SaaS cloud challenges. This will help SaaS CSPs and tenants to secure SaaS against internal and external threats.

RQ2:
What software security challenges face SaaS as identified in the GL?
This research question is designed to know the practitioner's opinions on SaaS security issues/challenges. These challenges will be identified by screening the GL available on the Google search engine.

RQ3: Which practices are suggested by the FL for improving SaaS cloud security?
This research question aims to analyze the selected primary studies to find the best practices that help in improving the security of the SaaS cloud.

RQ4
: Which practices are suggested by the GL for improving SaaS cloud security? This question will identify the best practices for improving SaaS security as suggested by practitioners in this area through the analysis of the identified GL.

RQ5
: Is there any similarity or discrepancy between R&P opinions regarding SaaS security issues and solutions?
This question will compare and analyze the results of the aforementioned four research questions to find the similarities and differences between R&P opinions.
The remaining paper organization is as follows: the background information is covered in Section 2. Section 3 summarizes the related work. The research methodology is defined in Section 4. The study's findings are presented in Section 5, which is accompanied by a review of the findings in Section 6. Section 7 concludes the paper by providing insights into future research. Appl. Sci. 2022, 12, 3953 4 of 29

Background
This section provides a general overview of some key terms related to CC with a special focus on SaaS security.

Cloud Computing (CC)
The cloud paradigm is not new. However, it is becoming more important to grasp the complexities of the CC language and principles as more organizations and businesses switch to cloud-based technology. According to a Cisco report, 94% of workloads will be managed by CC till 2021, compared to just 6% for conventional data centers [11]. NIST defines CC as a paradigm for providing global, useful, on-demand network access to a common pool of customizable computational resources that can be quickly provisioned and released with limited maintenance effort or intervention by service providers [12]. CC is different from traditional IT hosting systems as the user does not need to own the hardware. Rather, tenants only pay for the facilities they use [13].

CC Service Models
IaaS, PaaS, and SaaS are the three main service models for cloud solutions. IaaS provides customers with cloud-based storage, servers, networking, and other computing services. Although the user is still in charge of handling their programs, files, middleware, and so on, IaaS offers automated and flexible environments that give the user a lot of power and versatility. In PaaS, cloud users rent cloud-based services from service providers to build and deploy software. In other words, PaaS is a platform that makes developing, customizing, and deploying applications simpler and more effective. SaaS refers to software that is hosted, bundled, and distributed over the Internet by a third party. Enterprises will offload management and development costs to the provider by shipping business applications over the Internet. Email and customer relationship management tools are two common SaaS options [14].

Types of CC
PRC, PBC, and HYC are the three major cloud storage solutions. Each has its own set of benefits and drawbacks and which one a user (or company) selects will be determined by the nature of the data as well as the level of protection and management needed. A PBC is perhaps the most prominent form of CC. Both resources and supporting facilities are handled and accessed among many users off-site via the Internet (or tenants). A subscription service, such as Netflix or Hulu, is a clear example of a PBC at the individual user level. Rather than providing IT services to the public, the PRC delivers them to a small group of customers over the Internet or a private network. Various organizations adopt this option because it combines cloud mobility with greater customization and protection. Private and PBC elements are mixed in varying degrees in a HYC scheme. In relation to their freedom, the clouds in a hybrid environment operate together through an encrypted network, allowing data and applications to flow across them. This is a popular cloud solution because it provides companies with greater flexibility in fulfilling their IT needs [15,16].

SaaS Cloud Computing
SaaS is one of the cloud subscription service groups along with IaaS and PaaS. It allows businesses to access the programs they need without having to host them on their own servers. It has grown in popularity since it eliminates the need for organizations to buy servers and other resources, as well as retain an in-house support team. Instead, a SaaS provider hosts their applications and provides SaaS security and maintenance. Most business product providers, such as the Oracle Financials Cloud, also sell cloud implementations of their applications [17].

Benefits of SaaS
The global SaaS market is projected to expand at a rate of 21% annually over the next few years, hitting $117 billion by the end of 2022 [18]. The following key factors have contributed to the rise in popularity of SaaS.
Scalable and on-demand resources Quick implementation Easy maintenance and upgradation No staffing or infrastructure cost

SaaS Usage
In a number of cases, SaaS might be the best alternative [19], including 1.
Startups and small businesses that need to open an ecommerce site immediately and do not have the resources to deal with server or device problems.

2.
Short-term projects that necessitate short, simple, and cost-effective collaboration.

3.
Tax applications, for example, is an example of an application that is not used very much.

4.
Applications that include connectivity to both network and mobile devices

SaaS Security
SaaS is a widely used CC service model where security is the major issue due to its high dependence on third parties compared to other CC service models (see Figure 3). SaaS Security refers to the defense of user privacy and corporate data in subscription-based cloud applications. SaaS applications store a vast amount of personal data that can be accessed by many users from almost any device, putting privacy and critical data at risk. A large collection of enterprise software and data were hosted on in-house servers until a few years ago. From a security standpoint, this places the whole burden of proof on the operation, but at the very least, it is obvious what needs to be covered, and how. However, as more companies embraced SaaS tools, this led to increased security concerns. Since SaaS tools are hosted in the cloud, they raise new security issues, such as susceptibility to new malware and phishing attacks, as well as the risk of client data being exposed. Businesses can protect these cloud-based programs with the right SaaS Security software and by following best security practices. As a result, a thorough review of potential security issues/challenges, as well as best practices for enhancing SaaS security, is needed [20][21][22]. Appl. Sci. 2022, 12, x FOR PEER REVIEW 6 of 28 Figure 3. Comparison between three CC service models [23].

Literature Review
Before moving forward with the MVLR, it is essential to include a review of the existing studies to understand the current state of research. This section provides a quick rundown of some current research on SaaS security issues and mitigation strategies.
Hoener performed a systematic literature review (SLR) to gather SaaS security issues and solution. This SLR identified CC security requirements from scientific publications between January 2011 and March 2013. The identified requirements were categorized into Figure 3. Comparison between three CC service models [23]. Appl. Sci. 2022, 12, 3953 6 of 29

Literature Review
Before moving forward with the MVLR, it is essential to include a review of the existing studies to understand the current state of research. This section provides a quick rundown of some current research on SaaS security issues and mitigation strategies.
Hoener performed a systematic literature review (SLR) to gather SaaS security issues and solution. This SLR identified CC security requirements from scientific publications between January 2011 and March 2013. The identified requirements were categorized into a framework to assess their frequency. This research study also identified challenges in requirement assessment and proposed a solution for it [24]. Hashizume et al. presented the security issues of three cloud service models, IaaS, PaaS, and SaaS. The main security issues in CC, as stated in this article, are storage, virtualization, and networks. This paper also makes a distinction between vulnerabilities and threats, emphasizing the importance of comprehending these problems and establishing a connection between threats and vulnerabilities to determine which vulnerabilities lead to the implementation of these threats to render the framework more resilient. To mitigate these risks, several existing solutions were also identified. According to the findings of this paper, traditional protection systems may not function well in cloud environments because they constitute a dynamic architecture made up of a variety of technologies [25].
A systematic mapping study (SMPS) was performed by Juárez and Cedilloin in [26] to find the security issues of mobile computing. Based on the interpretation and assessment of 83 primary studies, this mapping study breaks down the security sub-characteristics from the ISO/IEC 25010 and compares them to the details contained in this study. The findings of the study indicate that there are sufficient studies to address the issue of confidentiality and integrity. However, accountability and non-repudiation need more attention.
A SMPS was performed by De Silva et al. to identify the security threats in CC. The aim of this study was to compile a list of the most recent publications in the literature that addressed security threats in CC. Centered on the Cloud Protection Alliance's "Top Threats to CC guide", this analysis presents metrics regarding existing research publications that deal with some of the seven security threats in CC. Furthermore, this research identified the most researched challenges, spreading the findings through 15 security domains, and identified the types of threats and solutions suggested. In view of these findings, the research focuses on publications that deal with meeting a regulatory requirement [27].
According to Zhou et al., users' use of CC technologies and applications is hampered by security and privacy concerns. The security and privacy issues raised by several CC system providers were investigated in this study. According to the results of the study, these considerations are insufficient. To meet the five goals (confidentiality, data integrity, availability, monitoring, and audit), more security techniques should be deployed in the cloud world, and privacy acts should be updated to adapt a new partnership between CSPs and CT in the cloud literature. According to the study, the CC literature needs to flourish until the protection and privacy problems are addressed [28].
Shankarwar and Pawar conduct a study of security and privacy problems as well as potential remedies. They discuss the benefits and disadvantages of the current methods to fully resolve the protection and privacy issues in the cloud world, as well as the advantages and drawbacks of existing methods to completely resolve the security and privacy issues [29]. According to Hussein and Khalid, security is still a major concern in the CC paradigm. User confidential data loss, data leakage, and the disclosure of confidential data are a few of the major security concerns. This paper presents a thorough review of the current literature on CC security issues and solutions [30].
A survey of the various security threats that pose a danger to the cloud is discussed by Kumbhar et al. This paper presents a survey focused on the various security problems that have arisen because of the nature of cloud infrastructure service distribution models. The paper discusses the security issues of three service models of CC, namely IaaS, PaaS, and SaaS in detail and provides recommendations to mitigate the mentioned threats. This paper examines CC security problems and categorizes them into different groups depending on the category of security. Furthermore, multiple trust-based solutions are classified according to how they provide trust in a collaborative setting. The results of the paper show that trust-based approaches for the cloud exist, but this does not explicitly solve all the discussed security issues [31].

Motivations for the Study
The preceding discussion demonstrates that security is a significant barrier to cloud adoption, especially in the case of SaaS clouds, where the tenant is completely reliant on the CSP for data security and control. As stated in the literature review, the existing literature has highlighted numerous security issues related to CC, but there is no explicit SLR, particularly MVLR, that provides the current SOTA for SaaS security issues and solutions. To bridge the gap and to provide a detailed overview of SaaS security for R&P, we conduct this MVLR. The detail of the MVLR is presented in subsequent sections.
The aim of this MVLR is to systematically review the literature on the SOTA and state of the practice (SOTP) of SaaS security. Both peer-reviewed and non-peer-reviewed literature are included in an MVLR (i.e., a form of SLR) [32,33]. In software engineering (SE), the SLR has become the most common way to conduct a literature review [34]. SLR only considers scientific contributions and excludes the GL. Since it lacks a vast volume of information produced by SE professionals, an SLR cannot always include an existing discipline of expertise. As a result, MVLRs are drawing more focus [33,[35][36][37]. We believe that an MVLR would be more beneficial than an SLR in the field of SaaS cloud security since there is a wide body of non-peer reviewed literature published by practitioners.

Contribution of the Study
In this MVLR, the basic issues and best practices for SaaS security are discussed. We looked at the current best practices to identify how to address SaaS security concerns. The following are the MLR's key contributions:

1.
It offers a taxonomy of various aspects of security issues that need to be addressed.

2.
It identifies best practices for improving SaaS security.

Research Methodology
The SLR and MVLR guidelines reported in the work in [34] were used to help with the framework for this MVLR. Our MVLR consists of three main phases (as shown in Figure 4). In phase-1, we develop the MVLR protocol, data extraction and analysis were undertaken in phase-2, and phase 3 provides the results, which are shown in a separate section. In the following, we discuss each phase in detail.

MVLR Protocol Development
The first phase of the MVLR is to develop an MVLR protocol. This phase consists of the following steps:

Research Identification
We identified the literature by employing a search strategy focused on the five research questions listed in the paper's introduction. The aim of these questions is to gain a deeper understanding of SaaS cloud security issues and challenges and identify best practices to mitigate SaaS security issues.

Search Strategy
The following section illustrates the search strategy that was used to acquire the relevant literature from multiple sources.

Sources of Data Collection
Our analysis covers the peer-reviewed FL and GL that was identified using manual and automated searches of related databases. Initially, we conducted a manual search Appl. Sci. 2022, 12, 3953 8 of 29 on Google Scholar to obtain an overview of the most recent literature and to make sure that enough literature exists in the area under research to conduct an MVLR. Further, the aim of this initial search is to compile a list of primary studies that may be used for the validation of the search string. We retrieved 10 primary studies that are closely related to the posed research questions. This preliminary literature review reveals that SaaS security is a major research issue, with numerous research studies in this field. However, the problem persists, and security remains a major barrier. Further, to the best of our knowledge, there is no existing MVLR on this topic. Therefore, there is a need to provide the current-SOTA picture of the SaaS security issues/challenges and best practices. In the second phase, an automated search was performed on six libraries: IEEE explorer, ACM, Science Direct, Springer, Wiley Online, and Google Scholar to retrieve the peer-reviewed literature. In the automated search, we used the advanced search option to match the search string with the title, abstract and keywords of the papers published between Jan 2010 to Jan 2021. For the GL search, we used the Google search engine similar to other MVLRs [35,36]. Since the Google search engine's algorithm retrieves and displays the most important results in the first few pages [34,36], we found the first 10 pages to be adequate for finding the most relevant literature. For example, the Google search retrieved 187, 000 records for the term "SaaS security" in January 2021, however relevant content was found only in the first 10 pages.

Search String
To ensure a robust search through several databases, we generated a search string. For the academic literature, we generated a search string focused on: (a) The keywords gathered from the primary studies (b) Synonym and alternative words used for the identified terms (c) Using the logical operator AND or OR to combine these terms.
We ran several pilot searches and refined our search string to make sure that all primary studies are retrieved by applying the search string. The search string was also tailored for different libraries. In the following, we describe each part of the search string.
("SaaS" OR "software as a service" OR "Software-as-a-Service") AND ("security" OR safety OR integrity OR confidentiality OR availability) AND (issues OR challenges OR problems OR limitations) AND (Practices OR guidelines OR recommendations OR checklist)).
We used "Software-as-a-service AND security" to find the GL and applied this string on the Google search.

Eligibility Criteria
We defined a set of inclusion and exclusion (I&E) criteria to choose the articles, as discussed in the following section. We used a narrow I&E criterion since this study is a mix of scientific FL and GL.

Inclusion Criteria for FL
The following points were considered for the inclusion of a FL.

•
Articles are written in English and the complete text is available.

•
Articles that focus on SaaS cloud security issues or challenges and solutions Exclusion Criteria for FL • Articles whose subject matter has nothing to do with SaaS security.

•
Articles not written in English.

Inclusion Criteria for GL
All the websites, blogs, white papers, news pages, or articles that discuss SaaS security challenges or practices to mitigate SaaS security issues were included. Appl. Sci. 2022, 12, x FOR PEER REVIEW 10 of 28 Exclusion Criteria for GL The exclusion criteria for GL were the same as for FL. We did not consider literature that was not written in English or that was not related to SaaS security.

Quality Assessment (QA)
To evaluate the strength of FL and GL, we defined the separate quality assessment criteria for both types of literature. We discuss these criteria separately in the following.

Quality Assessment of FL
A QA checklist was prepared to ensure the strength of the extracted FL. This checklist includes the following points. Exclusion Criteria for GL The exclusion criteria for GL were the same as for FL. We did not consider literature that was not written in English or that was not related to SaaS security.

Quality Assessment (QA)
To evaluate the strength of FL and GL, we defined the separate quality assessment criteria for both types of literature. We discuss these criteria separately in the following.

Quality Assessment of FL
A QA checklist was prepared to ensure the strength of the extracted FL. This checklist includes the following points.

•
The paper was cited by how many people.
The answer to this question was on a 3-point Likert scale, comprising yes, no and partial. If a paper was cited by more than 5 authors, the answer was yes, whereas if it was cited by between 1 and 5, then the answer was partial. Otherwise, the answer was no.

•
Was the paper accepted in a relevant journal, or not?
The answer to this question was yes/no and it was validated by reading the journal/conference list of topics.

Quality Assessment of GL
The quality of the GL was evaluated based on the following questions.

•
Is the publishing house a respectable one? • Does the individual author belong to a reputable organization? • Does the author have expertise in SaaS? Figure 5 shows the details of the studies retrieved from different sources after applying the I&E criteria. Our search retrieved 58 scientific papers from 6 different libraries and 35 grey studies retrieved from the Google search engine (as shown in Appendices A and B). Appl. Sci. 2022, 12, x FOR PEER REVIEW 11 of 28 • The paper was cited by how many people.
The answer to this question was on a 3-point Likert scale, comprising yes, no and partial. If a paper was cited by more than 5 authors, the answer was yes, whereas if it was cited by between 1 and 5, then the answer was partial. Otherwise, the answer was no.

•
Was the paper accepted in a relevant journal, or not?
The answer to this question was yes/no and it was validated by reading the journal/conference list of topics.

Quality Assessment of GL
The quality of the GL was evaluated based on the following questions.

Data Extraction, Synthesis and Analysis
We read, evaluated, extracted data, and synthesized the results from the selected papers based on the pre-defined RQs mentioned in the introduction to this paper. We extracted relevant data from each of the chosen sources that we used to address the research questions using a pre-defined data extraction form. We also gathered some general information about the papers, such as the authors' names, the country of publication, the venue of publication (journal/conference/workshop etc.), and the year it was published. Before determining how to collect the necessary data, we performed a pilot study on a selection of ten sources.

Results
We reviewed and categorized all the FL and GL separately and together in this section to determine the existing SOTA and SOTP. Next, we compared the views of the scholars and practitioners to identify the variations and parallels.

FL Analysis
This section analyzes the extracted scientific studies based on different criteria, such as the venue of publication, type of publication, year of publication, established SaaS security problems and challenges, and security best practices. The details of scientific literature retrieved from six different sources is given in Figure 6. Appl. Sci. 2022, 12, x FOR PEER REVIEW 12 of 2

Data Extraction, Synthesis and Analysis
We read, evaluated, extracted data, and synthesized the results from the selected pa pers based on the pre-defined RQs mentioned in the introduction to this paper. We ex tracted relevant data from each of the chosen sources that we used to address the researc questions using a pre-defined data extraction form. We also gathered some general infor mation about the papers, such as the authors' names, the country of publication, the venu of publication (journal/conference/workshop etc.), and the year it was published. Befor determining how to collect the necessary data, we performed a pilot study on a selectio of ten sources.

Results
We reviewed and categorized all the FL and GL separately and together in this sec tion to determine the existing SOTA and SOTP. Next, we compared the views of the scho ars and practitioners to identify the variations and parallels.

FL Analysis
This section analyzes the extracted scientific studies based on different criteria, such as the venue of publication, type of publication, year of publication, established SaaS security problems and challenges, and security best practices. The details of scientific literature retrieved from six different sources is given in Figure 6. The distribution of studies by the venue is shown in Table 1. All the retrieved re search articles are published in three venues only: journals, conferences, and workshops We can deduce the following facts from the statistics in Table 1.  The distribution of studies by the venue is shown in Table 1. All the retrieved research articles are published in three venues only: journals, conferences, and workshops. We can deduce the following facts from the statistics in Table 1.

•
Most of the retrieved studies that address SaaS security issues and best practices are published either in journals or conferences. There were only two studies published in workshops. One was published in the 2014 IEEE Globecom Workshop (GC Wkshps), while the other was published in IoTNAT'2016. • IEEE, ACM, and Springer have more conference publications.

•
Wiley has more journal publications. The author's affiliation was used to determine and rate the most involved countries in the field of SaaS security analysis. The aim of this ranking was to determine which countries' researchers are concentrating their efforts on this topic the most. If a paper had several authors, the country of the first author was selected. The demographic information of the retrieved studies is shown in Figure 7. According to Figure 7, India is the country which has published the most articles on SaaS security issues and solutions with a frequency of 23 out of 58 which is almost 40% of the total studies retrieved. The next is the USA with six out of 58 publications in the area under study, which is almost 10%. Publications from the UK were third in the pool with 4 out of 58, which is almost 7% of the total retrieved studies. Authors from Malaysia, China, and Greece each published three out of 58 studies. All the remaining studies were published by various other countries with a frequency of two and one. Appl. Sci. 2022, 12, x FOR PEER REVIEW 13 of 28

. Demographic Analysis of Retrieved Scientific Studies
The author's affiliation was used to determine and rate the most involved countries in the field of SaaS security analysis. The aim of this ranking was to determine which countries' researchers are concentrating their efforts on this topic the most. If a paper had several authors, the country of the first author was selected. The demographic information of the retrieved studies is shown in Figure 7. According to Figure 7, India is the country which has published the most articles on SaaS security issues and solutions with a frequency of 23 out of 58 which is almost 40% of the total studies retrieved. The next is the USA with six out of 58 publications in the area under study, which is almost 10%. Publications from the UK were third in the pool with 4 out of 58, which is almost 7% of the total retrieved studies. Authors from Malaysia, China, and Greece each published three out of 58 studies. All the remaining studies were published by various other countries with a frequency of two and one.   Figure 8 shows that SaaS security issues have been resolved to some extent with the advancement of the latest technologies and suitable security measures. However, some issues are still reported, which is why researchers are continuing to work in this area. The reasons for the curve's decline can be numerous. For example, it is possible that FL and GL on SaaS security exist but do not meet our I & E criteria. Second, with time, SaaS vendors are becoming aware of the security breaches and are taking suitable measures to secure their services to satisfy SaaS users. Third, there are several SaaS vendors in the market and security is one of the important concerns to be considered by users while selecting SaaS. Therefore, SaaS vendors have started paying more attention to it. Appl. Sci. 2022, 12, x FOR PEER REVIEW 14 of 28 issues are still reported, which is why researchers are continuing to work in this area. The reasons for the curve's decline can be numerous. For example, it is possible that FL and GL on SaaS security exist but do not meet our I & E criteria. Second, with time, SaaS vendors are becoming aware of the security breaches and are taking suitable measures to secure their services to satisfy SaaS users. Third, there are several SaaS vendors in the market and security is one of the important concerns to be considered by users while selecting SaaS. Therefore, SaaS vendors have started paying more attention to it.

SaaS cloud Security Issues and Challenges
We selected 58 scientific studies as primary studies after applying the I&E criteria. SaaS security issues and challenges were identified after a detailed screening of these primary studies. According to the studies' results, data loss/leakage is the key SaaS security challenge with a frequency of 41. Identity and access management and a lack of user control are the second most important challenges with a frequency of 39 each, after which is logical storage segregation and multi-tenancy/data locality with a frequency of 36. The remaining issues and challenges along with their frequency of occurrence and reference of scientific studies are detailed in Table 2. Table 2. SaaS security issues and challenges identified from FL.

SaaS cloud Security Issues and Challenges
We selected 58 scientific studies as primary studies after applying the I&E criteria. SaaS security issues and challenges were identified after a detailed screening of these primary studies. According to the studies' results, data loss/leakage is the key SaaS security challenge with a frequency of 41. Identity and access management and a lack of user control are the second most important challenges with a frequency of 39 each, after which is logical storage segregation and multi-tenancy/data locality with a frequency of 36. The remaining issues and challenges along with their frequency of occurrence and reference of scientific studies are detailed in Table 2. Table 2. SaaS security issues and challenges identified from FL.

Others 3
The issues or challenges with a frequency of one are combined under the heading 'others', including not incorporating security into SDLC, documentation, and different service delivery/receiving models.
The aim of identifying these SaaS security challenges and issues is to address our research question one (RQ1). This will help SaaS customers and service providers to gain an in-depth overview of SaaS security issues. Addressing these challenges will help to improve the security of the SaaS cloud.

SaaS Security Best Practices
We also analyzed the primary studies to extract SaaS security best practices to enhance the awareness of SaaS R&P about the SOTA. The best practices for improving SaaS security are detailed in Table 3. According to the results in Table 3, up-to-date security controls/standards represent the practice that is suggested by most researchers with a frequency of 39. The second-most mentioned practice for maintaining SaaS cloud security as suggested by the researchers is the use of strong encryption techniques with a frequency of 38. The third-most suggested practices for improving SaaS security are regulatory compliance/SLA compliance and multifactor authentication with a frequency of 24 each. The remaining best practices along with their frequency of occurrence and the studies in which they appeared are detailed in Table 3.
The best practices which were discussed in a single study only (with a frequency of one) were categorized under the heading 'others', including governments should keep their information assurance architectures secure and confidential, conduct service integrity test, interoperability management, and service conformity.

GL Analysis
This section analyzes the GL to obtain the practitioners' opinions about SaaS security issues and solutions. The selected GL is discussed based on different criteria, such as venue of publication, type of publication, year of publication, established SaaS security problems and challenges, and security best practices.

GL Analysis w.r.t Venue of Publication
To identify the GL related to SaaS security issues and best practices, we used the Google search engine. We applied the search string "SaaS security" on the Google search engine. The retrieved records belong to various categories, including white papers, blogs, websites, news, etc. The Google search engine retrieved thousands of records corresponding to our search string, but the relevant records were found in the first 10 pages only. We also checked the remaining few pages, but either the records were repeating, or they were not related to our search domains. The total number of selected GL was 35, as shown in Figure 9. These 35 records include 13 websites, nine blogs, eight reports, and five white papers. Table 3. SaaS security best practices as identified from FL.

Demographic Analysis of GL
While extracting the GL, we also collected demographic information (country in which the literature was published) to identify from which area practitioners are more active in providing their practical experience about SaaS security issues and solutions. Most of the websites and articles which were selected as GL were published in the USA, while only a few were published in other countries, including Canada, UK, India, Netherlands, and Estonia. This shows that practitioners from the USA are more concerned about CC security, especially SaaS security. However, the key aim of analyzing the GL was to extract SaaS security issues/challenges and best practices, and as some of the GL, especially websites, do not pertain to a specific year, we did not present this information.

SaaS Security Issues/Challenges Identified from GL
This section aims to answer RQ2 by identifying practitioners' opinions on SaaS security issues. According to the obtained results, data breaches/leakages is a key challenge faced by SaaS clouds with a frequency of 21, followed by identity and access management with a frequency of 17. Loss of control/visibility, an inability to maintain regulatory compliance, and an inability to monitor data in transit to and from cloud applications constituted the third-most frequently discussed challenge facing practitioners with a frequency of 10 each. The remaining challenges, frequency of occurrence, and the GL in which it was discussed are detailed in Table 4. Table 4. SaaS security issues and challenges (practitioners' opinions).

Best Practices for Improving SaaS Cloud Security (Practitioners' Opinions)
The selected GL was evaluated to extract suitable solutions/best practices to mitigate SaaS security issues and improve security. This will provide the practitioners' opinions regarding SaaS security. The extracted results are shown in Table 5. According to the results in Table 5, 26 out of 35 practitioners emphasize data encryption to ensure the security of SaaS cloud data, while governance and regulatory/standard compliance audits were suggested by 22 out of 35 practitioners. The third commonly agreed solution for improving SaaS security was the use of backups/recovery. The remaining best practices/solutions are detailed in Table 5 along with their frequency of occurrence and the corresponding GL reference. Table 5. SaaS security solutions/best practices (practitioners' opinion).

Best Practices Freq Occurrence in GL Reference
Multifactor authentication  18  GL1, GL2, GL3, GL4, GL6, GL9, GL10, GL11, G14, GL19 Some best practices were only suggested by a single practitioner. Hence, we combined all the practices with a frequency one into the category of 'others'. The practices which fall in the category of others include: watch for OWASP's top security issues, being careful with deadlines, and make security a priority.

Discussion
This MVLR has examined the pertinent factors that motivate the need to recognize SaaS security problems/challenges and the best practices that can resolve these security issues. We identified the current security challenges and best practices for improving SaaS cloud security issues during this study. Security improvements are being more widely recognized to accelerate SaaS cloud adoption. R&P from all over the world have been working for years to identify SaaS cloud security problems and challenges, as well as practices that can help solve these issues. However, the problem still persists, and challenges are reported from time to time. Therefore, this area has significant potential for innovation and research. Before proceeding to provide more solutions to existing security issues, there is a need to synthesize R&P opinions at a single place to provide the complete and current picture of the situation. To fill this gap, this MVLR has compiled a list of possible challenges from the FL and GL to raise the awareness of the cloud tenants as well as CSPs. Further, we also compiled practices for improving the security of the SaaS cloud both from the GL and peer-reviewed literature. This will help R&P address these issues and improve their SaaS cloud performance and adaption. In the following, we discuss these results in the light of the posed research questions.

RQ1: What software security challenges are involved in SaaS as identified in the FL?
This research question aimed to identify the SaaS security issues and challenges discussed by researchers in this area. A total of 58 studies were selected as primary studies after applying I&E criteria. When these studies were analyzed to find the SaaS security issues or challenges, about 18 challenges were identified from 58 primary studies. The frequency of the identified challenges was also calculated to understand the severity of each challenge. According to our results, the challenges with a higher frequency were: data security/data loss or leakage, identity and access management issues, lack of user control/visibility, logical storage segregation & multi-tenancy/data locality, insecure interfaces and APIs, and governance/regulatory compliance/SLA compliance with a frequency of 41, 39, 39, 36, 31, and 30, respectively. Network security/shared technology, virtualization issues/cloud & CSP migration issues, and malicious insiders were also key challenges with a frequency of 29, 27, and 20, respectively. Other challenges are also detailed in Table 2. The data in Table 2 provide a detailed overview of the SaaS security issues and challenges. This study will help R&P to obtain an overview of possible security issues and challenges. Based on these identified challenges, organizations can evaluate their current security breaches and can find timely solutions for these issues.

RQ2
: What software security challenges are involved in SaaS as identified in the GL?
Many researchers have identified general SaaS security issues from the FL. However, we did not find any study that has compiled R&P opinions together in a single study. To fill this gap and to provide a detailed overview of possible SaaS security issues as identified by academia and industry, this MVLR also extracted SaaS security issued from the GL. A total of 35 studies were selected as the GL, which include white papers, blogs, websites, and reports. According to our findings, the key security challenges identified by practitioners include data breaches/leakage, identity and access management, loss of control/visibility, inability to maintain regulatory/standard compliance and inability to monitor data in transit to and from cloud applications with a frequency of 21, 17, 10, 10, and 10 respectively. The detailed results are presented in Table 4. This will help SaaS tenants and CSPs obtain an overview of the practitioners' opinions on SaaS security issues and challenges.

RQ3: Which practices are suggested by the FL for improving SaaS cloud security?
This research question aims to analyze the selected primary studies to find the best practices that help in improving the security of the SaaS cloud. The mere identification of SaaS issues or challenges is not enough. Therefore, we extracted the solutions/best practices from these studies. Hence, 58 primary studies were also evaluated to find the security best practices, with a total of 14 practices identified from the peer-reviewed FL which provides a detailed overview of academia regarding SaaS security improvement. According to our findings, the researchers consider up-to-date security controls/standards, the use of strong encryption techniques, regulatory compliance/SLA compliance, multifactor authentication, and better enterprise infrastructure/proper data isolation as key practices for improving security with a frequency of 39, 38, 24, 24, and 23, respectively. A detailed list of these practices is given in Table 3. The results in Table 3 will help SaaS tenants and CSPs in relation to security analysis and improvement.
RQ4: Which practices are suggested by the GL for improving SaaS cloud security?
The experience of the people who are working in the industry is very useful in making any decision. To improve SaaS security, we also collected the opinions of practitioners who are actually working with SaaS and identified a list of practices mentioned by these practitioners. A total of 35 studies from the GL were evaluated and about 14 practices were identified from these studies that may help in improving SaaS security. According to practitioners, the most useful practices (based on frequency of occurrence) are data protection and encryption, governance and regulatory/standard compliance, and backups/recovery with a frequency of 26,22, and 20 respectively. The complete list of practices identified from the GL is given in Table 5. The data in Table 5 will help SaaS tenants and CSPs to evaluate their current security structure and will help them in relation to further improvement.

RQ5
: Is there any similarity or discrepancy between R&P opinions regarding SaaS security issues and solutions?
To compare academia and industry views regarding SaaS security issues and solutions, we analyzed the data presented in Tables 2-5. The security issues identified by academia and industry were almost same with a varying frequency of occurrence. Tables 6 and 7 provide the similarities and differences between R&P opinions regarding SaaS security issues /challenges and best practices, respectively.  The results of Table 6 shows that R&P are almost on the same pace regarding most of the issues/challenges facing SaaS security, as shown by the data in column 3 of Table 6.
The results of Table 7 shows that R&P are on the same pace regarding the best practices for SaaS security improvement. This shows that the provided MVLR is very helpful in identifying SaaS security challenges and solutions. According to the results of Table 7, SaaS security can be computed as where SS denotes SaaS security, X i refers to the security best practices mentioned in FL only, Y j refers to the security best practices mentioned in GL only, and Z k refers to the security best practices mentioned both in FL and GL. The coefficient 2 with Z k shows the dual weightage of these practices as both researchers and practitioners are in agreement about them.
To calculate the security ∑ 4 i=1 X i , we need to assign individual weightage to each practice according to its importance to the organization. The general calculation will be computed as where w 1 . . . w 4 are weights assigned to different practices, in the same way ∑ 10 k=1 Z k = w 1 z 1 + w 2 z 2 + w 3 z 3 + w 4 z 4 + w 5 z 5 + w 6 z 6 + w 7 z 7 + w 8 z 8 + w 9 z 9 + w 10 z 10 (4) The total security will be measured as

Study Implications for Both Academic and Industry
The findings of this MVLR will help SaaS tenants and CSPs to better understand the SaaS security challenges in detail. Taking benefit of this study, SaaS users will also come to know the most suitable solutions for addressing SaaS security challenges. The combined opinions of R&P will further strengthen their belief in the identified security challenges and solutions. The study also identifies the similarities and differences between SOTA and SOTP. This implies that SaaS CSPs can make real-time decisions to improve their security and SaaS adaptability. SaaS customers/tenants can also benefit from this study and play their part in improving SaaS security.

Research Limitations
There are some possible limitations to this research study. Since SaaS security problems and solutions constitute a complex paradigm with several terms, the search string used to find relevant articles could have missed some related terms. The study team established the I&E criteria that were used to evaluate and select the studies that were analyzed. The emphasis of this MVLR is not on an in-depth discussion of the reported solutions' limitations. We recommend that readers use the conclusions from this article with the aforementioned shortcomings in mind. Furthermore, some of the identified security concerns do not entirely meet the security standards of certain organizations.

Conclusions and Future Work
This paper provides the details of an MVLR that was conducted to better understand R&P opinions on SaaS security issues/challenges and solutions to improve SaaS cloud security. A total of 93 studies were extracted from the period from January 2010 to January 2021, including 58 scientific studies and 35 grey studies. The studies were retrieved using a search string and the studies which were included in this MVLR were extracted based on defined I&E criteria. First, we evaluated the scientific studies and provided a demographic analysis of these studies to identify the countries in which researchers are more active in research in this area. We also identified the key venues for publications in this area and in which years more studies were published. However, the key aim of the MVLR was to extract security issues/challenges and best practices to provide a complete SaaS security guide for SaaS tenants and CSPs. A total of 75 security issues and 44 best practices were extracted from 58 scientific studies, which are listed in Tables 2 and 3. The GL was also analyzed to obtain practitioners' opinions on SaaS security challenges and solutions. A total of 55 security issues/challenges and 47 best practices were identified from 35 grey studies. This MVLR provides a broad picture of the possible SaaS security issues and solutions for SaaS users. Further, SaaS vendors can evaluate their current security measures in the light of the mentioned security challenges and best practices. In the future, we plan to extend our research by proposing a standard solution to address SaaS security based on the issues and challenges identified in this MVLR. Furthermore, we are planning to map the identified challenges with best practices in order to provide deeper insights to the SaaS vendors so that they can take appropriate measures when they encounter any security challenges.