Applications and Evaluations of Bio-Inspired Approaches in Cloud Security: A Review

Cloud computing gained much popularity in the recent past due to its many internet-based services related to data, application, operating system, and eliminating the need for central hardware access. Many of the challenges associated with cloud computing can be specified as network load, security intrusion, authentication, biometric identification, and information leakage. Numerous algorithms have been proposed and evaluated to solve those challenges. Among those, bio-inspired algorithms such as Evolutionary, Swarm, Immune, and Neural algorithms are the most prominent ones which are developed based on nature’s ecosystems. Bio-inspired algorithms’ adaptability allows many researchers and practitioners to utilize them to solve many security-related cloud computing issues. This paper aims to explore previous research, recent studies, challenges, and scope for further analysis of cloud security. Therefore, this study provides an overview of bio-inspired algorithms application and evaluations, taking into account cloud security challenges, such as Identity and Authentication, Access Control Systems, Protocol and Network Security, Trust Management, Intrusion Detection, Virtualization, and Forensic.

such as Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and Software as a Service (SaaS) [5] as shown in Figure 1. Security is a crucial aspect of everyday computing, and as such, cloud computing is vehemently related to security due to sensitive and vital data stored on the cloud. Recently, lots of research is ongoing to answer the challenges of Big Data in cloud computing. Some of these challenges are related to information security, privacy, regulations, and performance issues. Apart from these, the risk of malicious insiders and the failure of cloud services are drawing attention to the consumer and retailers.
According to the Synopsis Cloud Security threat report (2018), 53% of organizations survey confirmed they have insider threats in their cloud system [6]. The culprit for the Marriott chain hotel data breach incident was an insider threat [7]. In addition, Application Programming Interface (API) in security is becoming more critical as it is the universal gateway for users to interact with cloud services. In 2018, British Airways hacked due to the instability of API within their cloud infrastructure [8]. Apart from this, other attacks like malware attack, cross-cloud attack, side-channel attacks are also rising at large.
The current age of Cyber Warfare makes big data and cloud computing a lucrative target for adversaries worldwide. A vast number of applications, services, and data storage transfers within the cloud are increasing due to convenience facilities (i.e., numerous storage, loss prevention, mobility), even though they do not provide enough security.

A. SECURITY CHALLENGES IN CLOUD-COMPUTING
Security in big data has three essential viewpoints: information assurance, security protocols, and data protection [9].
Security management for distributed computing aims to solve big data management issues, preserve the integrity of systems, and protect cyberspace from threats [10]. Big data security focuses on real-time dynamic security observations to identify any potential threat/vulnerability or even unusual behaviors. While keeping the data access speed at a feasible level, there is always a risk of confidential information leakages. As big data heavily relies on cloud computing, cloud computing security aspects are required to be thoroughly evaluated. The concern behind this is that cloud computing security is not the same as traditional computing security [11]. There are several layers in a cloud computing network, such as wireless network distribution, peer-to-peer systems, and virtualization platforms [12]. Hence, just protecting different blocks may not be enough to secure the systems entirely. After the full system starts to work, unique security issues might arise in the individual blocks. Also, security challenges associated with small parts of cloud computing from the mobile browser, web engine to Linux server kernel, arise at an alarming rate. Cloud computing inherits them, which makes the security of cloud computing more challenging (see Table 1). It seems integral for security providers to come up with new solutions to protect the viability/usability of cloud systems. Different types of algorithms are employed to protect the cloud, such as Encryption, Sobol sequence, Stripping algorithms, and Biologically inspired algorithms [13]. Researchers who are working with innovative algorithms such as different machine learning algorithms or other cryptographic computations to face off the challenges are considering nature-inspired algorithms in their research. Table 2 presents different domains and factors in cloud computing and proposed mechanisms to mitigate the challenges associated with cloud security using different algorithms. This paper intends to provide an overview of the application and evaluation of bio-inspired algorithms-Evolutionary, Swarm, Immune, and Neural Network-in cloud security, taking into account published literature from 2010 to March 2020. The study outcomes expect to serve as a subsidy for current and future researchers, practitioners, and the general public to understand the overall trend, importance, and limitations of bio-inspired algorithms.
The rest of the paper is organized as follows: section II describes different bio-inspired algorithms, Section III provides an overview of past recent research trends, Section IV reviews the current work in cloud security, Section V defines the challenges, opportunities, and future research areas in the improvement of bio-inspired algorithms. Finally, Section VI draws research conclusions and future directions in cloud security.

II. BIO-INSPIRED ALGORITHMS
Living systems exist in nature, follow some systematic procedures to fulfill their needs, and it is possible to define that procedure mathematically. These procedures inspired mathematicians to develop some algorithms, and these algorithms are now known as the biologically inspired algorithm. In general, biologically inspired algorithms can be classified into four sections, such as Evolutionary algorithms, Swarm algorithms, Immune algorithms, and Neural algorithms [13].
A. EVOLUTIONARY ALGORITHMS Evolutionary algorithms are inspired by the evolution theory, proposed by Charles Darwin [27]. It is a population-based algorithm that finds the best result using biological phenomena such as reproduction, mutation, recombination, and selection. Some prominent examples of Evolutionary algorithms are Genetic Programming (GP), Gene Expression programming (GEP), and the Strength Pareto Evolutionary Algorithm. GP mainly works by encoded computer problem solutions known as the population to gene like structure and try to find out the best by using different evolutionary techniques. The behavior of DNA-RNA replication inspires the development of GEP. It works as a genotype-phenotype system, where a genome transfers genetic information, and phenotype adapts with the environment. The application of Evolutionary algorithms as a means of cloud security can be observed in Access Control Systems [14], Protocol Network Security [15], and Trust Management [16].

B. SWARM ALGORITHMS
Some insects or animals' behavior has inspired researchers to solve many problems in science and engineering (i.e., swarms of bees, the flock of birds). Swarm Intelligence (SI), a sub-field of artificial intelligence, is inspired by biological swarms' intelligent behavior and solves real-world problems by simulating such natural actions. The algorithms are postulated from the collective knowledge, observed in nature, such as a group of birds or fish's movement or how they behave as a singular unit. It has several characteristics, such as adaption, scalability, speed, autonomy, parallelism, and fault tolerance [28]. Some well-known examples are the Ant System, Ant Colony System, and Bacterial Foraging Optimization algorithm. Ant Colony algorithm is based on ants' behaviors-randomly searching for food and finding the optimal solution to return the food to its colony; the Bees algorithm is slightly different as the search is not random in the beginning. On the other hand, Bacterial Foraging Optimization is another inspiring example of SI, where bacteria's formation based on environmental parameters were inspired to develop a sophisticated algorithm for multi-agent optimization. The implementation of Swarm algorithms in cloud security can be found in the following fields: Authentications [17], [18], Forensics [19], and Virtualizations [20].

C. IMMUNE ALGORITHMS
Immune systems of living beings become an example of sophisticated defense algorithms. For example, the human VOLUME 8, 2020 body can use its prior knowledge of harmful bacteria to prepare for future defense. The Immune algorithms are adaptive and best for developing a dynamic defense algorithm for network security and privacy. Some examples of these algorithms are the Clonal and Negative Selection, Artificial Immune Recognition, Immune Network, and Dendritic Cell algorithm. A detailed explanation regarding these algorithms can be found in [29]. The Clonal Selection algorithm is best known for optimization and in the area of pattern analysis. A Negative Selection algorithm is widely used for anomaly detection; it prepares standard and unseen features, making it better suited, especially in defense techniques where the attacker's presence is unknown. Similarly, the Dendritic Cell algorithm is another example of an Immune algorithm that works in musicale and layered. Few applications in cloud security domains that use Immune algorithms include Identity and Authentication [21], [22], Protocol and Network Security [23].  [30]. Figure 2 demonstrates the evolution of nature-inspired algorithms over time in cloud security. Most of the research related to security in cloud computing is based on metaheuristics, while current research emphasizes hyper-meta heuristic.

III. ANALYSIS OF RESEARCH TRENDS
To represent the extent of a literature survey of nature-inspired algorithms in cybersecurity, we developed a four step framework: (i) web-based database search and review, (ii) reference analysis, (iii) abstract probe, and (iv) entire text review. To identify the potential research paper for literature review, we used different keywords to search for articles on the online database, such as Google Scholar, the Web of Science, IEEE, and Springer. The keywords used during the screening process were cybersecurity, Evolutionary algorithm, Swarm algorithm, Immune algorithm, and Neural algorithm. Relevant published articles from 2010 to June 2020 are included in this paper, but some essential papers from different years, published before 2010 are also included due to their significant contribution to this field/scope of work. The screening process yielded more than 7000 papers related to bio inspired algorithms (among them around 1127 was related to cybersecurity). After careful review based on the number of citations, the relevance of cloud security, and publication on a reputable journal (i.e., IEEE, Springer), more than 100 papers have been taken into account in this literature. Table 3 shows the number of papers published in bio-inspired fields until 2020 on different domains.
From Table 3, it is evident that more than half of the referenced literature considered Neural Networks (NN). Swarm algorithms seemed less prevalent in security research. However, some notable and productive work were observed in the Access Control Systems (ACS) and Intrusion Detection (ID) with Swarm Intelligence (SI). Nearly half of the Immune algorithm in security research has been done in Intrusion Detection. NN dominates the chart when it comes to security sectors like Forensic areas. Table 3 also revealed that ACS and ID research is the predominant area of focus currently. Also, a decent amount of literature considered non-traditional fields like Forensics and Virtualization in their study. On the other hand, identity and authentication rely merely on traditional cryptography, where research and development focus seem minimal. And, Trust Management is relatively new, and not much progress has made over the years.
In conclusion, the utilization of bio-inspired algorithms in cloud security-related research are neck to neck for Immune and Evolutionary, Swarms are way behind, and Neural algorithms are far ahead. However, further assessment is needed to evaluate those evidence regarding security in cloud computing, which is briefly in detail in section IV.

IV. LITERATURE REVIEW A. IDENTITY AND AUTHENTICATION SECURITY
Identity and Authentication Security (IAS) are significant concerns for cloud computing, and most IAS rely on cryptographic performances [31] or other computational complexity [32]. As shown in Figure 3, there are many layers in cloud computing and ensuring security related issues such as confidentiality, integrity, and data availability is hard to establish on each layer [24]. Bio-inspired algorithms are used by many researchers to tackle those issues. For instance, [33] presented a new security architecture for user identification that includes two-factor authentication. Their method proposed to keep login data, and encryption/decryption in one database, and rest uploaded accessories on different databases. They noted that this approach would stop any harmful or corrupted files uploaded by hackers or attackers in a cloud system.
In IAS, most security-related works have been constructed using NN algorithms, and more often applied to Keystroke and face identifications [17], [18], [24], [34]. For example, Wei et al. (2011) [19] briefly analyzed NN in password authentication systems; similarly, [35] developed password verification techniques for multi-server architecture using Neural algorithms. However, authentication research using SI did not explore that much. Still, some of the notable works that draw attention is as follows: the author(s) in [17] used Particle Swarm for palm and face identity; [18] used Immune algorithms for signature identification; [21], [22], [36], [37] developed a negative authentication system using Negative Selection algorithms. Apart from this, some studies also distilled promising results for the authentication process considering the GA based approach with adaptive selections [34], [38].

B. ACCESS CONTROL SYSTEMS
The Access Control Systems (ACS) are restricted to digital assets based on user privilege. In the cloud computing system, asset management and maintenance are the most vulnerable side of a security breach. Probably, the reason behind the most biology-inspired algorithms concentration on ACS. As cloud systems work with ever-expanding data growth, ACS requires to be adaptive. The adaptive nature of Evolutionary algorithms inspired security researchers to design ACS using GA. For example, Yadav et al. (2020) developed a secure ACS cloud computing, an integrated DNA Morse code-base systems considering collision, man-in-themiddle, and internal attacks [14].
Some notable works tried to use cloud users' attributes to design effective access control systems [40], [41]. Among all the nature-inspired algorithms, Swarm algorithms introduced by Di Caro et al. (1999) [42], gained much popularity Two new network functions designed to attenuate the impact of authentication traffic, generated by secondary authentication on the 5G home network, such as Secondary Authentication Function (SAF) and Authentication Data Management Function (ADMF). Primarily, the secondary authentication operates by SAF process. The SAF gets requests from the user, processes the request, and interacts with ADMF to increase the audibility of communique environment [39].
for developing security in the ACS [43]- [46]. For example, A study conducted by [46] proposed the Mutual Trust-Based Access Control model (MTBAC) in cloud computing. [44] introduced SI based systems for routing in mobile networks. [45] uses Ant Colony Clustering and Linear Genetic Programming in web-based mining. In ACS, Immune algorithms are mostly used to solve resource allocation [47], [48] and the hierarchical key design for access control [49], [50]. Apart from this, some literature considered NN in ACS designs [51], [52]. For instance, [51] uses Feed-Forward NN; similarly, [52] uses NN to optimize resources and classify new resources for parental control systems. Figure 4 displayed an encryption-based systemthe data owner provides encryption keys for the user to access stored encrypted data-in a cloud server.

C. PROTOCOL AND NETWORK SECURITY
Merging different layers of cloud computing requires to develop adequate Protocol and Network Security (PNS).
Biology inspired algorithms are also widely used to design faster network routing and effective security protocols. For instance, [53] and [54] aims to develop wireless network routing using Evolutionary algorithms. Bashkar et al. (2014) [15] introduced a notable work by securing cluster-based data aggregation in Wireless Sensor Networks (WSN) using GA. Sing et al. (2014) [55] also developed a mobile ad-hoc network using GA.
These works inspired several researchers to design more network protocols using adaptive natures [56], [57]. Most recent works are now aiming to improve the quality service on routing protocols [23], [58], [59]. Rathee et al. (2019) [60] proposed an Ant Colony Optimization based WSN system for energy balancing in secure routing. They have considered network lifetime, QoS, and security as the primary factor during the experiment. They showed that their proposed model QEBSR performed better than the other two existing models: distributed energy balanced routing model and energy-efficient routing model. [56], [57] used the ACO to develop WSN and routing protocols, and later improved by  [64], where the Immune algorithm was applied to improve the lifetime and stability period of WSN. Meanwhile, Neural Networks (NN) algorithms are also thoroughly used in this field. NN mostly used to improve the hopping and optimizing the network route [65]- [67].

D. TRUST MANAGEMENT
Trust Management (TM) is becoming more and more important as social media content starts to dominate cloud computing. Nevertheless, still, the amount of research in this field is insufficient. Service level agreement, recommender, and reputation-based models are some of the trust models in cloud security, as shown in Figure 5. Several studies stated promising results in the last few years by using Evolutionary and Immune algorithms in cloud computing [16], [68], [69]. Tahta et al. (2015) [16] developed a peer to peer systems for TM using GA. Tau et al. (2006) [68] improved trustworthiness using Immune algorithms (IA); it inspires more researchers to use IA in Trust Management [69], [70]. [71] introduced an anti-attack TM scheme for the Vehicular Ad-hoc network, an example of adaptive forgetting element-based strategies that develop trustworthy VAN networks by avoiding malicious vehicles and assisting with trusted vehicles. [72] proposed an optimized trust-aware recommender system using GA. In this paper, the author developed a model for choosing the most suitable nodes for the skeleton of recommender searching. Using this method, they were able to reduce the skeleton's maintenance cost more than 90%. On the other hand, the NN is also used to develop a trust model. However, most of its notable works are regarded to classify or optimize data/networks' reputations. For example, [73] developed a distributed network systems, which later proved helpful in cloud computing [74]. Even though the NN-based approach is used extensively in cloud computing due to its advantage in trust management, the NN models have some disadvantages: lower accuracy [75], unable to handle local optimal and convergence [76] problems, and fail to process multi-class attacks [77]. Therefore, continuous evolution is required.

E. INTRUSION DETECTION
Intrusion Detection (ID), one of the most researched topics in cyber-attack, is divided into three categories-Specificationbased, Anomaly-based, and Signature-based-as shown in Figure 6. Some bio-inspired algorithms, such as Immune and Neural algorithms, are proven to be useful to detect intrusion due to their optimized classification techniques. [78] describes the application of Immune-based algorithms in ID systems; [79] presents a case study to tailor the bio-inspired algorithm for ID. On the other hand, Wang et al. (2018) [80] proposed an improved Immune algorithm for industrial cloud storage. Alaparthy et al. (2018) [81] analyzed a multi-level ID system based on the Immune algorithm. Many studies address the application of ANN in ID. For example, Tran et al. [82] Proposed ANN-based adaptive boosting and probabilistic methods; Alarcon et al. [83] introduced Adopted Recurrent Neural Networks for low false alerts and better accuracy. However, one of the most significant drawbacks of their proposed method is that it requires high processing power and is also unable to detect insider attacks.
Some researchers also attempt to use the Genetic Algorithm (GA), but it was found that Immune algorithms were more effective than GA to Intrusion Detection [84]- [86]. Nonetheless, several studies have used the Neural net over the years [87]- [89]. For instance, in 2018, [89]   Previously, Privacy issues were categorized as Identity and Authentication issues or a matter of ACS and managed by Access Control. As the internet age grows, Privacy issues become more complicated and contemplated as an individual pivotal threat in cloud computing. Several techniques were proposed to tackle those threats, using the nature-inspired algorithm (i.e., Neural Network, Evolution, and Immune); for instance, Yuan et al. (2014) introduced an NN-based algorithm using back-propagation for preserving privacy in cloud computing [92]. Said et al. (2018) developed a Clustering Coefficient based Genetic Algorithm (CC-GA) for detecting communities in social networks [93].
As social networks' applications are continuously growing, it becomes necessary to analyze the network communities' structure and their potential features in terms of privacy concerns. Besides, large scale social networks also need excessive storage space, which leads to high computation cost. Considering the risk of social user privacy leakage in the clustering process, Bian et al. (2019) proposed a Markov Clustering algorithm (DP-MCL) with different privacy premise [94]. The algorithm (DP-MCL) stores the social network as an input in the adjacency matrix. It compresses the network sizes to minimize the network scale based on different privacy and guaranteed accurate clustering; however, they did not consider reducing that vast social network in an efficient and timely manner during the study.
When the data set becomes more abundant, it is common for users to store their data in a cloud more often. However, to ensure better security, data stored as an encrypted form. However, if those data have multiple owners, then problem raises as it becomes difficult to store as an encrypted form due to the different keys while it is also necessary to keep the communication cost minimum for the user's satisfaction. Resolving those issues, Li et al. (2017) [95] proposed a deep learning approach for multi-key privacy-preserving in cloud computing. The algorithm combines double decryption and Fully Homomorphic Encryption (FHE) to improve the communication facility and minimum cost. However, they did not test their model performance in a real-world scenario; also, the cost reduction was not significant.

G. VIRTUALIZATION
Virtualization is a system that supports distributing an individual physical instance of an asset or an application between many consumers. Different digital technology, such as desktop, storage, and application, can be virtualized through cloud systems as shown Figure 7. Most of the cloud Virtualization problem is resource allocation problem, also known as NP-Hard; and, finding exact solutions is complicated for large scale data [96]. In virtual security, the most widely used biological inspired algorithm is the Evolutionary algorithm.  [80] proposed a PSO based framework to solve the VMP problem by extending original PSO to discrete searching space. [100] proposed an energy efficient algorithm using a heuristic algorithm, called MinPR for virtual machine placement optimization (considered power consumption and resource efficiency). Using MinPR, authors were able to prioritize the power efficient ones over all the active physical machines, to reduce resource wastage by maximizing and balancing resource utilization. However, they did not consider the dependency between VMs and the data center network topology. [101] proposed an Ant Colony Optimization (ACO) heuristic algorithm to host the VMs into PMs. They performed a local search with ACO that significantly improved the solution by minimizing the number of PMs. Cao (2019) [102] proposed a multi-objective GA. Their goal was to minimize energy consumption and communication traffic, which caused performance bottlenecks. In addition, Usman (2019) [96] proposed a new approach to reduce resource wastage and increase energy efficiency for VM allocation using Flower Pollination in the cloud data-center. The distribution employs a strategy called Dynamic Switching Probability (DSP). Using DSP frameworks, they were able to find the optimal solution quickly and balance the exploration of the global search and the local search. Their proposed method outperformed Genetic-Algorithms for Power-Aware (GAPA) by 21.8%, the Order of Exchange Migration (OEM) Ant Colony system by 21.5%, and First-Fit Decreasing (FFD) by 24.9%. However, their proposed method did not consider the multi-objective approach of E-FPA to consolidate the data center resource, which may need to be investigated. Also, the optimal solution is not highly scalable. Besides, Khurana and Singh (2019) [103] also uses FPA based algorithms along with GWO to improve VM efficiency. To conduct the experiment, they have considered the following parameters: the number of tasks, the number of workflows, the number of VM, the MIPS, and the number of processors. However, their method has a high computational cost.
The author in [104] proposed a method for VM placement by OH-BAC algorithm. They have considered the following parameters for their study: load balance, CPU utilization, memory, bandwidth, storage size, and memory. Using OH-BAC techniques, they were able to find the optimal PM with the least power consumption. However, their proposed method needs more Service Level Agreement (SLA) time per Active Host (SLATAH).
Another study conducted by Liu et al. [105] formulated VMP with a reliability model and analyzed its complexity with an approximation algorithm. Their proposed model proved to be effective and efficient in solving traffic-aware and reliability guaranteed VMP problems. However, they did not consider some VM related challenges such as VM backups or VM migration during their experiment.   [106] studied the influence of five different parameters to develop an optimized virtual machine. The examined factors were broker cost, time duration, bandwidth, ram speed, and overall cost. The authors used the Honey Bee approaches for balancing load across the virtual machines and were able to maximize the throughput. However, during this experiment, some of the essential factors, such as server CPU power and memory, which also plays a vital role in VM efficiency, were not considered.
Many past recent studies considered data security as the most significant security challenge in cloud computing. [107] proposed a Firefly Swarm approach for developing new connections in social networks based on big data analysis. [20] offered Bacterial Foraging to prevent security threats on the flow of big-data information. [108] developed a novel hybrid bio-inspired algorithm using a Multilayer Perceptron (MLP) to handle big data security. However, even with significant advantages, bio-inspired algorithms (such as Bacterial Foraging, Swarm approach, and MLP) are often not suitable for scalability-vulnerable considering fault tolerance, and agility.

H. FORENSICS
Cloud Forensics is the combination of traditional computer Forensics, small-scale digital device Forensics, and network Forensics. Cloud Forensics usage includes troubleshooting, log monitoring, due diligence, data, and system-recovery, regulatory compliance, investigation on digital crimes, civil cases, and policy violations as illustrated in Figure 8. Cloud Forensics becomes popular in law enforcement in order to understand and track criminal activity by gathering information from digital devices like smartphones, computers, and smart sensors. Over the years, biology-inspired algorithms have not been explored much in this field. Additionally, no specific or substantial techniques have been developed to serve as a guide for cloud technology. Therefore, current tools and techniques are insufficient to succeed in proper Forensics reports due to the lack of adequate training and compatibility issues [3]. Mukkamala et al. (2003) [26] proposed a Neural Network approach to identify the significant features for network Forensics analysis.
Many security challenges associated with cloud Forensics includes false alert and unanimous attack in the systems. To deal with these issues, more often, the bio-inspired based approach is utilized. For example, [109] proposed an NN based technique; [110] suggested a combined kernel PCA and GA to reduce training time and [111] developed a fuzzy logic and ANN-based techniques for anomaly detection.
Forensics identification can be classified as fingerprint, facial recognition, log analysis, and user data analysis. Law enforcement agencies more often use facial recognition techniques to identify person or crime investigation [112]. While using existing algorithms, it is possible to achieve up to 100% accuracy; for Forensics identification, it is not that easy. The challenges faced by this discipline consist of several factors, and among them, the quality of the image itself is the biggest problem. For example, most of the CCTV's digital video recorder (DVR) is normally kept at resolution 720 pixel wide with H.264 compression. After some time, the video resolution was sampled at a smaller size to accommodate more recording space. Beyond that, signal noise, color noise, illumination problems also cause many dead ends to Forensics analysis. To deal with these issues, several studies have proposed a solution, based on a bio-inspired algorithm [19], [112], [113]. [19] proposed Particle Swarm Optimization (PSO) method associated with Support Vector Machine (SVM), named by PSO-SVM for facial recognition in cloud forensic. Even though proposed PSO-SVM methods hypothesized good accuracy, the performance deteriorated with random values while velocity is calculated. To overcome that issue, [112] proposed a modified feature extraction method named as AAPSO-SVM.
Here, researchers showed that their proposed method performed better than traditional bio-inspired algorithms such as ABC, PSO, and PSO-SVM, and achieved up to 85% accuracy. Another study conducted by [113] compares three existing techniques proposed by the previous work, such as PSO-SVM, AAPSO-SVM, and OPSO-SVM [19], [112]. Their result illustrated that AAPSO-SVM performed better compared to other algorithms. Even though authors in [113] found AAPSO-SVM as the best algorithm compared to the other two, [114] suggested that PSO-SVM based method should be used in classification stage-as it performed better than other algorithms when using with AAM (feature extraction techniques).
Most of the cloud Forensics relates to virtual evidence like facial recognition. However, fingerprint is also considered as cloud Forensics, even though it is physical evidence. We did not consider fingerprints as cloud security related issues during this literature. However, in fingerprint detection, several studies showed promising results using bio-inspired algorithms. For example, [115], achieved 90% accuracy using PSO-SVM.
Still, cloud Forensics faces several issues such as tenancy, integrity, privacy, and Encryption. Addressing those issues, Pandi et al. (2020) found the lacking of advanced tools as the main culprit behind the progress of Forensics analysis in VOLUME 8, 2020 the current environment. [116]. On the other hand, existing techniques fail to reduce leaking confidential informationdocuments, images, and videos-of victims from end devices like smartphones and tabs [117]. Several studies often suggest difficulty accessing the evidence using logs as another hinder in cloud Forensics investigation [118], [119]. [120] argued that the problem could be solved if logs through the eucalyptus cloud environment. In general, issues, like data fragment [119], lack of trust issues [121], and cloud infrastructure isolation [4], are still some of the challenges faced by cloud Forensics till days. However, the typical limitations of all nature-inspired algorithms have to deal with time complexity issues [109]- [111]. Additional challenges associated with cloud Forensics are the unification of log formats, synchronization of timestamps, the exponential increase of digital devices accessing the cloud, and ineffective encryption key management [122] Table 4 lists a list of papers organized primarily based on the problem domain, security function, algorithm, and application. Most of the articles published between 2018 to 2020 were presented along with previous literature, and interested readers are recommended for further assessment based on their needs.

V. DISCUSSION AND POTENTIAL REMARKS
Based on the presented literature of security on cloud computing, this segment describes the general findings of the existing research and the future direction of cloud computing studies considering nature-inspired algorithms.
Most of the papers prioritize the following elements in their experiments: task scheduling [96], [123], [124], bio-metric identification [112], [125], network optimization [60], [126]- [129], and network stability [123], [130], [131]. ACO may be very promising when implemented to single-goal optimization, even though more common GA have somewhat overtaken. A few efforts have been made to regulate ACO to a multi-objective paradigm, which is an exciting avenue; this is perhaps deserve further study. In combination with optimization objective capabilities, the possibility of meta-heuristics might also improve network performance and require additional attention.
The lack of proper security, cloud computing is accessible for hackers/attackers to misuse this platform. Moreover, a model developed with existing algorithms such as Fuzzy Logic, Swarm Intelligence, and Neural Network is still unable to secure the network entirely. Additionally, with big data, it is becoming difficult for the traditional VOLUME 8, 2020 algorithm to extract features, especially in cloud Forensics, or to optimize the model, mostly for customer satisfaction. Thus, it is necessary to introduce a more sophisticated and advanced strategy to develop a secure network. Compared to the traditional approach, sophisticated algorithms, like deep learning, ANN, or CNN based methods, gained lots of popularity because of their better accuracy on large datasets. As a result, recently, researchers are using a deep learning-based approach constantly to address cybersecurity issues. However, one of the significant drawbacks of those algorithms is that they are time-consuming and need to find a better way to handle the time complexity issues. Additionally, researchers may also consider optimizing the routing time, authentication time, and network cost for further study.

VI. CONCLUSION
In conclusion, this research survey provided a comprehensive analysis of the latest research techniques and algorithms related to biologically inspired algorithms used in cloud computing. The referenced literature mainly focused on two bio-inspired algorithms: PSO and NN approach to tackling maximum cloud computing security-related problems. Additionally, Algorithms like the Ant Colony, Fruitfly, and Grasshopper drew interest to decipher specific issues among researchers. However, some of the studies also developed a secured network system using the GA method. Most of the results were simulation-based and did not integrate with another matching, learning, forecasting models; therefore, a potential future application might be interesting considering multi-objective optimization using GA based approach. Finally, we would like to suggest some of the possible scopes, shortly for researcher and practitioner as a brainstorming concept: reducing the reaction time and maximizing VM's resource allocation considering the QoS factor; improving the load stability in WSN using RCNN learning; SVM-PSO based community Forensics and RNN techniques for Intrusion Detection.   ABBAS Z. KOUZANI (Member, IEEE) received the B.Sc. degree in computer engineering from the Sharif University of Technology, Iran, the M.Eng. degree in electrical and electronic engineering from The University of Adelaide, Australia, and the Ph.D. degree in electrical and electronic engineering from Flinders University, Australia. He was a Lecturer with the School of Engineering, Deakin University, Australia, and then a Senior Lecturer with the School of Electrical Engineering and Computer Science, University of Newcastle, Australia. He is currently a Professor with the School of Engineering, Deakin University. He is also the Director of the Deakin University's Advanced Integrated Microsystems (AIM) research group. He provides research leadership in embedded, connected, and low-power devices, circuits, and instruments that incorporate sensing, actuation, control, wireless transmission, networking and IoT, data acquisition/storage/analysis, AI, energy harvesting, power management, and fabrication for tackling research questions relating to a variety of disciplines including healthcare, ecology, mining, infrastructure, automotive, manufacturing, energy, utilities, and agriculture. He has produced over 370 publications, including one book, 17 book chapters, 180 journal articles, and 181 fully refereed conference papers. He has three patents and two pending patents. He has been involved in over $15 million research grants, and has managed projects and delivered research solutions to over 25 Australian and International companies. He received several awards, including Outstanding Contribution to Scholarly Publication Award', School of Engineering, Deakin University, in 2019. He has supervised 24 research fellows/assistants, and produced 28 Ph.D. and six Masters by Research completions. He is also involved in supervision of 12 Ph.D. students.

M.
A. PARVEZ MAHMUD received the B.Sc. degree in electrical and electronic engineering and the master of engineering degree in mechatronics engineering. After the successful completion of his Ph.D. degree with multiple awards, he worked as a Postdoctoral Research Associate and Academic with the School of Engineering, Macquarie University, Sydney. He worked with the World University of Bangladesh (WUB) as a Lecturer for more than two years and with the Korea Institute of Machinery and Materials (KIMM) as a Researcher for about three years. He is currently an Alfred Deakin Postdoctoral Research Fellow with Deakin University. He has produced over 50 publications, including one authored book, three book chapters, 29 journal articles, and 21 fully refereed conference papers. He is also involved in the supervision of six Ph.D. students with Deakin University. His research interests include energy sustainability, secure energy trading, microgrid control and economic optimization, machine learning, data science, and micro/nanoscaled technologies for sensing and energy harvesting. He accumulated experience and expertise in machine learning, life cycle assessment, sustainability and economic analysis, materials engineering, microfabrication, and nanostructured energy materials to facilitate technological translation from the lab to real-world applications for the better society. He received several awards, including Macquarie University Highly Commended Excellence in Higher Degree Research Award 2019. He was involved in teaching engineering subjects in the electrical, biomedical and mechatronics engineering courses at the School of Engineering, Macquarie University, for more than two years. He is also a Key Member of the Deakin University's Advanced Integrated Microsystems (AIM) research group. Apart from this, he is actively involved with different professional organizations, including Engineers Australia and IEEE.