A Survey on Secure Cloud: Security and Privacy in Cloud Computing

Tables index

Veiw figure View Table

4.1. Security Issues with SaaS

SaaS provides application services on demand such as email, conferencing software, and business applications such as ERP, CRM, and SCM ^[39]. SaaS users have less control over security among the three fundamental delivery models in the cloud. The adoption of SaaS applications may raise some security concerns.

Data Security: Data security is a common concern for any technology, but it becomes a major challenge when SaaS users have to rely on their providers for proper security ^{[40, 41, 42]}. Data security includes the specific controls and technologies used to enforce information governance. This has been broken out into three sections to cover detection of data migration to cloud, protecting data in transit to the cloud and between different providers and protecting data once it’s within the cloud. The SaaS provider is the one responsible for the security of the data while is being processed and stored ^[30]. In SaaS, organizational data is often processed in plaintext and stored in the cloud. Also, data backup is a critical aspect in order to facilitate recovery in case of disaster, but it introduces security concerns as well ^[41]. Also cloud providers can subcontract other services such as backup from third-party service providers, which may raise concerns. Moreover, most compliance standards do not envision compliance with regulations in a world of Cloud Computing ^[40]. In SaaS model, the process of compliance is complex because data is located in the provider’s datacenters, which may introduce regulatory compliance issues such as data privacy, segregation, and security, that must be enforced by the provider.

Application Security: Since applications are typically delivered via the Internet through a Web browser ^{[40, 45]}. However, flaws in web applications may create vulnerabilities for the SaaS applications. Security challenges in SaaS applications are not different from any web application technology, but traditional security solutions do not effectively protect it from attacks, so new approaches are necessary ^[41]. Attackers have been using the web to compromise user’s computers and perform malicious activities such as steal sensitive data ^[43]. The Open Web Application Security Project (OWASP) has identified the ten most critical web applications security threats ^[46]. There are more security issues, but it is a good start for securing web applications.

Multi-Tenancy: The impact of multi-tenancy is visibility of residual data or trace of operations by other user or tenant. In this case, multiple consumers with same or different organization use same resources or applications. Information security is one of the prime factors for this phase. Since data from multiple tenants is likely to be stored in the same database, the risk of data leakage between these tenants is high. Security policies are needed to ensure that customer’s data are kept separate from other customers ^[47].

Access Control: Accessing applications over the internet via web browser makes access from any network device easier, including public computers and mobile devices. However, it also exposes the service to additional security risks. The Cloud Security Alliance ^[48] has released a document that describes the current state of mobile computing and the top threats in this area such as information stealing mobile malware, insecure networks (WiFi), vulnerabilities found in the device OS and official applications, insecure marketplaces, and proximity-based hacking.

4.2. Security Issues with PaaS

PaaS cloud (public or private) offers an integrated environment to design, develop, test, deploy, and support custom applications developed in the language the platform supports. PaaS application security comprises two software layers: Security of the PaaS platform itself (i.e., runtime engine), and Security of customer applications deployed on a PaaS platform ^[49]. PaaS providers are responsible for securing the platform software stack that includes the runtime engine that runs the customer applications.

Mashups combine more than one source element into a single integrated unit. Thus, PaaS models also inherit security issues related to mashups such as data and network security ^[50]. Also, PaaS users have to depend on both the security of web-hosted development tools and third-party services.

From the perspective of the application development, developers face the complexity of building secure applications that may be hosted in the cloud. However, developers also have to understand that any changes in PaaS components can compromise the security of their applications. Besides secure development techniques, developers need to be educated about data legal issues as well, so that data is not stored in inappropriate locations. Data may be stored on different places with different legal regimes that can compromise its privacy and security. In PaaS, developers do not usually have access to the underlying layers, so providers are responsible for securing the underlying infrastructure as well as the applications services ^[51]. Even when developers are in control of the security of their applications, they do not have the assurance that the development environment tools provided by a PaaS provider are secure.

Access Control: In the PaaS delivery model, the CSP is responsible for managing access control to the network, servers, and application platform infrastructure. However, the customer is responsible for access control to the applications deployed on a PaaS platform. Access control to applications manifests as end user access management, which includes provisioning and authentication of users.

Table 2. Security Monitoring for Cloud Service

Download as

Tables index

Veiw figure View Table

4.3. Security Issues with IaaS

With IaaS, cloud users have better control over the security compared to the other models as long there is no security hole in the virtual machine monitor ^[41]. They control the software running in their virtual machines, and they are responsible to configure security policies correctly ^[52]. However, the underlying compute, network, and storage infrastructure is controlled by cloud providers. IaaS provides a pool of resources such as servers, storage, networks, and other computing resources in the form of virtualized systems, which are accessed through the Internet ^[55]. IaaS providers must undertake a substantial effort to secure their systems in order to minimize these threats that result from creation, communication, monitoring, modification, and mobility ^[53]. Unlike PaaS and SaaS, IaaS customers are primarily responsible for securing the hosts provisioned in the cloud. Customers of IaaS have full access to the virtualized guest VMs that are hosted and isolated from each other by hypervisor technology. Hence customers are responsible for securing and ongoing security management of the guest VM. Some of the new host security threats in the public IaaS include:

• Attacking unpatched, vulnerable services listening on standard ports (e.g., FTP, NetBIOS, SSH)

• Hijacking accounts that are not properly secured (i.e., weak or no passwords for standard accounts)

• Stealing keys used to access and manage hosts (e.g., SSH private keys)

• Deploying Trojans embedded in the software component in the VM or within the VM image (the OS) itself

• Attacking systems that are not properly secured by host firewalls

Virtualization: It allows users to create, copy, share, migrate, and roll back virtual machines, which may allow them to run a variety of applications ^{[53, 54]}. However, it also introduces new opportunities for attackers because of the extra layer that must be secured ^[43]. Virtual machine security becomes as important as physical machine security, and any flaw in either one may affect the other ^[56]. Virtualized environments are vulnerable to all types of attacks for normal infrastructures; however, security is a greater challenge as virtualization adds more points of entry and more interconnection complexity ^[45].

Virtual Machine Monitor (VMM): VMM or hypervisor is responsible for virtual machines isolation; therefore, if the VMM is compromised, its virtual machines may potentially be compromised as well. The VMM is low-level software that controls and monitors its virtual machines, so as any traditional software it entails security flaws ^[57]. A vulnerable hypervisor could expose all user domains to malicious insiders. An attacker can compromise the migration module in the VMM and transfer a victim virtual machine to a malicious server. Keeping the VMM as simple and small as possible reduces the risk of security vulnerabilities, since it will be easier to find and fix any vulnerability. A malicious VM can infer some information about other VMs through shared memory or other shared resources without need of compromising the hypervisor ^[58]. Using covert channels, two VMs can communicate bypassing all the rules defined by the security module of the VMM ^[59]. Thus, a malicious Virtual Machine can monitor shared resources without being noticed by its VMM, so the attacker can infer some information about other virtual machines.

An attacker with a valid account can create an image containing malicious code such as a Trojan horse. If another customer uses this image, the virtual machine that this customer creates will be infected with the hidden malware. Moreover, unintentionally data leakage can be introduced by VM replication ^[60]. Some confidential information such as passwords or cryptographic keys can be recorded while an image is being created. If the image is not “cleaned”, this sensitive information can be exposed to other users.

Virtual Networks: Virtual Networks increase the VMs interconnectivity, an important security challenge in Cloud Computing ^[61]. The most secure way is to hook each VM with its host by using dedicated physical channels. However, most hypervisors use virtual networks to link VMs to communicate more directly and efficiently. For instance, most virtualization platforms such as Xen provide two ways to configure virtual networks: bridged and routed, but these techniques increase the possibility to perform some attacks such as sniffing and spoofing virtual network [57, 62]^{[57, 62]}.

Availability Management: Availability considerations for the IaaS delivery model should include both a computing and storage (persistent and ephemeral) infrastructure in the cloud. IaaS providers may also offer other services such as account management, a message queue service, an identity and authentication service, a database service, a billing service, and monitoring services. Hence, availability management should take into consideration all the services that user depends on for his IT and business needs.

5. Attacks on Cloud Data

Attacks on cloud data is increasing day by day with various techniques. Table 3 shows the idea about threats, attacking type with layer where it occurs mostly. A threat is a potential attack that may lead to a misuse of information or resources, and the term vulnerability refers to the flaws in a system that allows an attack to be successful. Some of them are:

Cross Site Scripting (XSS) attacks: Cross-site Scripting (XSS) refers to client-side code injection attack wherein an attacker can execute malicious scripts (also commonly referred to as a malicious payload) into a legitimate website or web application. XSS is amongst the most rampant of web application vulnerabilities and occurs when a web application makes use of un-validated or un-encoded user input within the output it generates. In order for an XSS attack to take place the vulnerable website needs to directly include user input in its pages. An attacker can then insert a string that will be used within the web page and treated as code by the victim’s browser ^[26]. A cross-site scripting vulnerability may be used by attackers to bypass access controls such as the same-origin policy. Cross-site scripting carried out on websites accounted for roughly 84% of all security vulnerabilities documented by Symantec as of 2007 ^[27]. It can be classified as: non-persistent and persistent.

In non-persistent, holes show up when the data provided by a web client, most commonly in HTTP query parameters or in HTML form submissions, is used immediately by server-side scripts to parse and display a page of results for and to that user, without properly sanitizing the request.

Persistent XSS occurs when the data provided by the attacker is saved by the server, and then permanently displayed on "normal" pages returned to other users in the course of regular browsing, without proper HTML escaping. A classic example of this is with online message boards where users are allowed to post HTML formatted messages for other users to read.

SQL injection attacks: Here malicious code is inserted into a standard SQL code. Thus the attackers gain unauthorized access to a database and are able to access sensitive information ^[28]. Sometimes the hacker’s input data is misunderstood by the web-site as the user data and allows it to be accessed by the SQL server and this lets the attacker to have know-how of the functioning of the website and make changes into that. Various techniques like: avoiding the usage of dynamically generated SQL in the code, using filtering techniques to sanitize the user input etc. are used to check the SQL injection attacks. A proxy based architecture towards preventing SQL Injection attacks which dynamically detects and extracts users’ inputs for suspected SQL control sequences has been proposed in ^[29].

Man in the Middle attacks (MITM): Another class of attacks, quite popular to SaaS, is termed as MITM. In such an attack, an entity tries to intrude in an ongoing conversation between a sender and a client to inject false information and to have knowledge of the important data transferred between them. Various tools implementing strong encryption technologies like: Dsniff, Cain, Ettercap, Wsniff, Airjack etc. have been developed in order to provide safeguard against them. A detailed study towards preventing man in the middle attacks has been presented in ^[30].

DoS/DDos Attacks: In computing, a denial-of-service (DoS) attack is an attempt to make a machine or network resource unavailable to its intended users, such as to temporarily or indefinitely interrupt or suspend services of a host connected to the Internet. A distributed denial-of-service (DDoS) is where the attack source is more than one–and often thousands of-unique IP addresses. Criminal perpetrators of DoS attacks often target sites or services hosted on high-profile web servers such as banks, credit card payment gateways. A denial-of-service attack is characterized by an explicit attempt by attackers to prevent legitimate users of a service from using that service. There are two general forms of DoS attacks: those that crash services and those that flood services. Permanent denial-of-service (PDoS), also known loosely as phlashing, ^[35] is an attack that damages a system so badly that it requires replacement or reinstallation of hardware. ^[36] Unlike the distributed denial-of-service attack, a PDoS attack exploits security flaws which allow remote administration on the management interfaces of the victim's hardware, such as routers, printers, or other networking hardware. The attacker uses these vulnerabilities to replace a device's firmware with a modified, corrupt, or defective firmware image—a process which when done legitimately is known as flashing. This therefore "bricks" the device, rendering it unusable for its original purpose until it can be repaired or replaced.

Table 3. Cloud Threads

Download as

Tables index

Veiw figure View Table

6. Cloud Security Controls

Cloud security architecture is effective only if the correct defensive implementations are in place. Efficient cloud security architecture should recognize the issues that will arise with security management ^[22]. The security management addresses these issues with security controls. These controls are put in place to safeguard any weaknesses in the system and reduce the effect of the attack. While there are many types of controls behind the cloud security architecture, they can usually be found in one of the following categories ^[22]:

• Preventive controls: Preventive controls strengthen the system against incidents, generally by reducing if not actually eliminating vulnerabilities. Strong authentication of cloud users, for instance, makes it less likely that unauthorized users can access cloud systems, and more likely that cloud users are positively identified.

• Detective controls: Detective controls are intended to detect and react appropriately to any incidents that occur. In the event of an attack, a detective control will signal the preventative or corrective controls to address the issue ^[22]. System and network security monitoring, including intrusion detection and prevention arrangements, are typically employed to detect attacks on cloud systems and the supporting communications infrastructure.

• Deterrent controls: Deterrent controls are intended to reduce attacks on a cloud system. Much like a warning sign on a fence or a property, deterrent controls typically reduce the threat level by informing potential attackers that there will be adverse consequences for them if they proceed.

• Corrective controls: Corrective controls reduce the consequences of an incident, normally by limiting the damage. They come into effect during or after an incident. Restoring system backups in order to rebuild a compromised system is an example of a corrective control.

7. Security Approach through Existing Technologies

Although there are many benefits to adopting Cloud Computing, there are also some significant barriers to adoption. One of the most significant barriers to adoption is security, followed by issues regarding compliance, privacy and legal matters ^[44]. Because Cloud Computing represents a relatively new computing model, there is a great deal of uncertainty about how security at all levels (e.g., network, host, application, and data levels) can be achieved and how applications security is moved to Cloud Computing [2, 65]^{[2, 65]}. That uncertainty has consistently led information executives to state that security is their number one concern with Cloud Computing ^[49]. Literature review is presented in this section that deals with existing cloud security models, methodology and algorithms.

Cloud-Trust - a Security Assessment Model for Infrastructure as a Service (IaaS) Clouds ^[66] is presented by Gonzales, D. et. all. It is a cloud architecture reference model that incorporates a wide range of security controls and best practices, and a cloud security assessment model – Cloud-Trust – that estimates high level security metrics to quantify the degree of confidentiality and integrity offered by a Cloud Computing Systems (CSS) or cloud service provider (CSP). Cloud-Trust is used to assess the security level of four multi-tenant IaaS cloud architectures equipped with alternative cloud security controls and to show the probability of CCS penetration (high value data compromise) is high if a minimal set of security controls are implemented. CCS penetration probability drops substantially if a cloud defense in depth security architecture is adopted that protects virtual machine (VM) images at rest, strengthens CSP and cloud tenant system administrator access controls, and which employs other network security controls to minimize cloud network surveillance and discovery of live VMs. In optimized fine-grained and fair pricing scheme ^[67], two tough issues are addressed for IaaS platform: (1) the profits of resource providers and customers often contradict mutually; (2) VM-maintenance overhead like startup cost is often too huge to be neglected.

Biometric encryption ^[5] is proposed to improve the confidentiality in Cloud computing for biometric data. The privacy of a particular user is an issue in biometric data i.e. the face reorganization data for a famous and important people. Also, this paper discussed virtualization for Cloud computing, as well as biometrics encryption.

In decentralized multi-authority attribute-based signature (DMA-ABS) scheme ^[68], no central authority and no trusted setup are required. The proposed DMA-ABS scheme for a large class of (non-monotone) predicates is fully secure (adaptive-predicate unforgivable and perfectly private) under a standard assumption, the decisional linear (DLIN) assumption, in the random oracle model. In Outsourced ABS ^[69], the computational overhead at user side is greatly reduced through outsourcing intensive computations to an untrusted signing-cloud service provider (S-CSP).

Multi-factor authentication (MFA) ^[10] is an approach to user validation that requires the presentation of two or more authentication factors. Given the popularity of cloud systems, MFA systems become vital in authenticating users. A privacy-preserving multi-factor authentication system utilizing the features of big data called MACA ^[10]. In MACA, the first factor is a password while the second factor is a hybrid profile of user behavior. The hybrid profile is based on users' integrated behavior, which includes both host-based characteristics and network flow-based features. MACA is the first MFA that considers both user privacy and usability combining big data features.

Homomorphic encryption is a form of encryption that allows computations to be carried out on ciphertext, thus generating an encrypted result which, when decrypted, matches the result of operations performed on the plaintext. Fully Homomorphic encryption (FHE) ^[31] allows straightforward computations on encrypted information, and also allows computing sum and product for the encrypted data without decryption. In addition the homomorphic property of various cryptosystems can be used to create many other secure systems, for example secure voting systems,^[32] collision-resistant hash functions, private information retrieval schemes, and many more. Craig Gentry ^[33] using lattice-based cryptography, described the first plausible construction for a fully homomorphic encryption scheme. Gentry's scheme supports both addition and multiplication operations on cipher texts, from which it is possible to construct circuits for performing arbitrary computation.

An ID-Based User Authentication Scheme for Cloud Computing ^[11] supports higher security levels and lower computation costs. An efficient authentication scheme for distributed mobile cloud computing services is proposed in ^[12]. The proposed scheme provides security and convenience for mobile users to access multiple mobile cloud computing services from multiple service providers using only a single private key. The security strength of the proposed scheme is based on bilinear pairing cryptosystem and dynamic nonce generation. Security Enhanced Anonymous Remote User Authentication and Key Agreement for Cloud Computing are proposed in ^[13]. It enables a user and a cloud server to authenticate each other anonymously and establish a secure channel between them. Thus, only the user and the cloud server may learn the messages exchanged and no entity except themselves can learn the real identities of the message senders.

A new scheme is proposed ^[14] for mutual authentication where the user and cloud server can authenticate one another. The protocol is designed in such a way that it uses steganography as an additional encryption scheme. The scheme achieves authentication using secret sharing. Secret sharing allows a part of the secret to be kept in both sides which when combined becomes the complete secret. The secret contains information about both parties involved. Further, out of band authentication has been used which provides additional security. The proposed protocol provides mutual authentication and session key establishment between the users and the cloud server. Also, the users have been given the flexibility to change the password.

7.1. ABE Schemes

Sahai and Waters proposed a new type of IBE – Fuzzy Identity-Based Encryption ^[7]. It is also known as Attribute-Based Encryption (ABE). In their work, an identity is viewed as a set of descriptive attributes. Different from the IBE, where the receiver could decrypt the message if and only if his identity is exactly the same as what specified by the sender, this fuzzy IBE enables the decryption if there are identity overlaps‘ exceeding a pre-set threshold between the one specified by sender and the one belongs to receiver. However, this kind of threshold-based scheme was limited for designing more general system because the threshold based semantic cannot express a general condition. In ABE, a user has a set of attributes in addition to its unique ID. It can be divided into two main categories: KP-ABE scheme and CP-ABE scheme.

7.2. KP-ABE Scheme

In Key-policy ABE or KP-ABE (Goyal et al. ^[8]), the sender has an access policy to encrypt data. Cipher-text is associated with a set of attributes, which partially represents the cipher-text‘s encryption policy. A writer whose attributes and keys have been revoked cannot write back stale information. The receiver receives attributes and secret keys from the attribute authority and is able to decrypt information if it has matching attributes.

Unfortunately, with a drawback that the access policy is built into the secret key, the data owner in a KP-ABE scheme cannot decide the one who can decrypt the cipher text, and he can only choose a set of attributes to control the access of cipher texts. Besides, the access structure is a monotonic access structure which cannot express the negative attribute to exclude the participants with whom the data owner does not want to share data. Subsequently, Ostrovsky et al. ^[70] proposed a scheme with a non-monotonic access structure where the secret keys are labeled with a set of attributes including positive and negative attributes. Comparatively, the ABE scheme with non-monotonic access structure can express a more complicated access policy. Unfortunately, this mechanism doubles the size of the cipher text and secret key and adds encryption/decryption overheads at the same time. Ostrovsky et al.’s initial construction is recently improved by Lewko et al. ^[71] who us a new technique to achieve user revocation and design the most efficient non-monotonic KP-ABE scheme. KP-ABE schemes, In this methodology the cipher text size grows linearly with the number of cipher text attributes and the only known exception only supports restricted forms of threshold access policies.

7.3. CP-ABE Scheme

In 2007, using a monotonic access tree as access structure, Bethencourt et al. ^[72] proposed the first CP-ABE construction. Their scheme can support flexible access control policies like the KP-ABE ^[8] scheme, but the security proof is in the generic group model. Cheung and Newport ^[73] provided a provably secure CP-ABE scheme which is proved to be secure under the standard model and their scheme supports AND gate on positive and negative attributes as its access policy. In 2011, Waters ^[9] proposed a new methodology for realizing CP-ABE under concrete and non-interactive cryptographic assumptions in the standard model. He expressed access control by a linear secret sharing scheme (LSSS) matrix over the attributes in the system (previously used structures can be expressed succinctly in terms of an LSSS). In this most efficient scheme, the cipher text size and the encryption/decryption overheads increase linearly with the complexity of the access formula. As a result, his scheme achieves the same performance and functionality as Bethencourt et al.’s ^[72]. Finally, Lewko et al. ^[74] recently leveraged the encoding technique from Waters’s scheme ^[9] to propose an ABE scheme that achieves adaptive (nonselective) security. Their scheme is based on the Composite order groups, which results in some loss of practical efficiency when compared with Water’s. Emura et al. ^[84] improved the efficiency and achieved hidden policies.

Table 4 shows the comparison among various existing CP-ABE based Encryption/Security systems.

Table 4. CP-ABE Technologies Comparison

Download as

Tables index

Veiw figure View Table

7.4. Dual-Policy ABE Scheme

In 2009, Attrapadung and Imai ^[75] presented a new ABE scheme called the Dual-Policy ABE. Basically, it is a conjunctively combined scheme of Goyal et al.’s KP-ABE scheme ^[8] and Waters’ CP-ABE scheme ^[9]. It allows simultaneously two access control mechanisms over encrypted data. One involves policies over objective attributes ascribed to data and the other involves policies over subjective attributes ascribed to user credentials. These two access control mechanisms can only allow either functionality above one at a time. What is more, the security proof is based on decisional bilinear Diffie-Hellman exponent (DBDHE) assumption.

7.5. MA-ABE Scheme

Multi-authority ABE schemes ^{[76, 77]} can be divided into two types. One needs a central authority (CA, for short) which is used to guarantee the proper decryption and can also decrypt all cipher texts, such as schemes ^{[76, 78]}, while the other does not need a CA, such as schemes ^{[79, 80]}.

Table 5. Comparison of computational cost

Download as

Tables index

Veiw figure View Table

Table 6. Working Idea Comparison for MA-ABE

Download as

Tables index

Veiw figure View Table