Problems and Solutions

Cloud computing security:

Key references to security guidance and top threats analysis highlight different security issues related to cloud computing that require further studies for being appropriately handled and, consequently, for enhancing technology acceptance and adoption. Emphasis is given to the distinction between services in the form of software (SaaS), platform (PaaS) and infrastructure (IaaS), which are commonly used as the fundamental basis for cloud service classification. However, no other methods are standardized or even employed to organize cloud computing security aspects apart from cloud deployment models, service types or traditional security models. Aiming to concentrate and organize information related to cloud security and to facilitate future studies, in this section we identify the main problems in the area and group them into a model composed of seven categories, based on the aforementioned references. Namely, the categories are: network security, interfaces, data security, virtualization, governance, compliance and legal issues. Each category includes several potential security problems, resulting in a classification with subdivisions that highlights the main issues identified in the base references:

I. Network security:

Problems associated with network communications and configurations regarding cloud computing infrastructures. The ideal network security solution is to have cloud services as an extension of customers’ existing internal networks, adopting the same protection measures and security precautions that are locally implemented and allowing them to extend local strategies to any remote resource or process.

(a) Transfer security: Distributed architectures, massive resource sharing and virtual machine (VM) instances synchronization imply more data in transit in the cloud, thus requiring VPN mechanisms for protecting the system against sniffing, spoofing, man-in-the-middle and side-channel attacks.

(b) Firewalling: Firewalls protect the provider’s internal cloud infrastructure against insiders and outsiders. They also enable VM isolation, fine-grained filtering for addresses and ports, prevention of Denial-of-Service (DoS) and detection of external security assessment procedures. Efforts for developing consistent firewall and similar security measures specific for cloud environments reveal the urge for adapting existing solutions for this new computing paradigm. (c) Security configuration: Configuration of protocols, systems and technologies to provide the required levels of security and privacy without compromising performance or efficiency.

II. Interfaces: Concentrates all issues related to user, administrative and programming interfaces for using and controlling clouds.

(a) API: Programming interfaces (essential to IaaS and PaaS) for accessing virtualized resources and systems must be protected in order to prevent malicious use.

(b) Administrative interface: Enables remote control of resources in an IaaS (VM management), development for PaaS (coding, deploying, testing) and application tools for SaaS (user access control, configurations).

(c) User interface: End-user interface for exploring provided resources and tools (the service itself), implying the need of adopting measures for securing the environment.

(d) Authentication: Mechanisms required enabling access to the cloud. Most services rely on regular accounts consequently being susceptible to a plethora of attacks whose consequences are boosted by multi-tenancy and resource sharing.

III. Data security: Protection of data in terms of confidentiality, availability and integrity (which can be applied not only to cloud environments, but any solution requiring basic security levels).

(a) Cryptography: Most employed practice to secure sensitive data, thoroughly required by industry, state and federal regulations.

(b) Redundancy: Essential to avoid data loss. Most business models rely on information technology for its core functionalities and processes and, thus, mission-critical data integrity and availability must be ensured.

(c) Disposal: Elementary data disposal techniques are insufficient and commonly referred as deletion. In the cloud, the complete destruction of data, including log references and hidden backup registries, is an important requirement.

IV. Virtualization: Isolation between VMs, hypervisor vulnerabilities and other problems associated to the use of virtualization technologies.

(a) Isolation: Although logically isolated, all VMs share the same hardware and consequently the same resources, allowing malicious entities to exploit data leaks and cross-VM attacks. The concept of isolation can also be applied to more fine-grained assets, such as computational resources, storage and memory.

(b) Hypervisor vulnerabilities: The hypervisor is the main software component of virtualization. Even though there are known security vulnerabilities for hypervisors, solutions are still scarce and often proprietary, demanding further studies to harden these security aspects.

(c) Data leakage: Exploit hypervisor vulnerabilities and lack of isolation controls in order to leak data from virtualized infrastructures, obtaining sensitive customer data and affecting confidentiality and integrity.

(d) VM identification: Lack of controls for identifying virtual machines that are being used for executing a specific process or for storing files.

(e) Cross-VM attacks: Includes attempts to estimate provider traffic rates in order to steal cryptographic keys and increase chances of VM placement attacks. One example consists in overlapping memory and storage regions initially dedicated to a single virtual machine, which also enables other isolation-related attacks.

V. Governance: Issues related to (losing) administrative and security controls in cloud computing solutions.

(a) Data control: Moving data to the cloud means losing control over redundancy, location, file systems and other relevant configurations.

(b) Security control: Loss of governance over security mechanisms and policies, as terms of use prohibit customer-side vulnerability assessment and penetration tests while insufficient Service Level Agreements (SLA) lead to security gaps.

(c) Lock-in: User potential dependency on a particular service provider due to lack of well-established standards (protocols and data formats), consequently becoming particularly vulnerable to migrations and service termination.

VI. Compliance: Includes requirements related to service availability and audit capabilities.

(a) Service Level Agreements (SLA): Mechanisms to ensure the required service availability and the basic security procedures to be adopted.

(b) Loss of service: Service outages are not exclusive to cloud environments but are more serious in this context due to the interconnections between services (e.g., a SaaS using virtualized infrastructures provided by an IaaS), as shown in many examples. This leads to the need of strong disaster recovery policies and provider recommendations to implement customer-side redundancy if applicable.

(c) Audit: Allows security and availability assessments to be performed by customers, providers and third-party participants. Transparent and efficient methodologies are necessary for continuously analyzing service conditions and are usually required by contracts or legal regulations. There are solutions being developed to address this problem by offering a transparent API for automated auditing and other useful functionalities.

(d) Service conformity: Related to how contractual obligations and overall service requirements are respected and offered based on the SLAs predefined and basic service and customer needs.

Numerous Cloud Identity Management Systems (IDMSs) have been proposed so far; however, most of those systems are neither widely accepted nor considered highly reliable due to their constraints in terms of scope, applicability and security. In order to achieve reliability and effectiveness in IDMs for Cloud, further extensive research needs to be carried out to critically examine Cloud based IDMSs and their level of security.

Another aspect of the Cloud systems is complexity. The problem in understanding cloud systems stems from the fact that it is simply quite difficult to model them. Cloud is a very dynamic system with numerous users, devices and networks, connecting and disconnecting simultaneously with the cloud. This complexity is to such an extent that it can perhaps be likened to the complexity of a human brain where neurons connect and change their synaptic structure continuously to store information.

However, what is a problem here is the fact that unlike the brain, where the connecting neurons are already authenticated, cloud systems require extensive authentication as well as identity management systems. Still, these are simply not enough to cater for the ever-growing requirements of novel paradigms such as the Internet of Things (IoT) in relation to its connectivity with the cloud.

- Cloud-computing standardization

The lack of standards could make cloud computing trickier to use. It could also restrict imple­mentation by limiting interoperability among cloud platforms and causing inconsistency in areas such as secu­rity and interoperability. For example, the lack of standardization could keep a customer trying to switch from a pri­vate to a public cloud from doing so as seamlessly as switching browsers or e-mail systems. In addition, it would keep users from knowing the basic capabilities they could expect from any cloud service.

Interoperability between offerings and the portability of services from one provider to another is very impor­tant to the customer to maximize the expected [return on investment] from cloud computing. A lack of security standards - addressing issues such as data privacy and encryption - is also hurting wider cloud-computing adoption.

Cloud standardization

Its most basic, cloud computing is simply the delivery of applications; security and other services; storage and other infrastructures; and plat­forms such as those for software development to users over the Internet or a private cloud.

Cloud computing appeals too many organizations because it minimizes the amount of hardware and software that users must own, maintain, and upgrade. In essence, users pay only for the computing capability they need.True interoperability requires translation of specific application and service functionality from one cloud to another, and this won’t happen without standardization.

A key standardization issue involves virtualization, which plays a critical role in most cloud-comput­ing approaches. Virtualization’s flexibility lets cloud providers optimize workloads among their hardware resources. This also enables users to, for exam­ple; connect to storage without having to know about server names and addresses, which would be the case in a traditional network.

In virtualization, hypervisors manage a host server’s processing and other resources so that it can run multiple virtual machines (VMs), using different operating systems and other platforms. Each cloud platform has its own type of hypervisor. Cloud systems utilizing different hypervisors won’t interoperate, in part because they don’t use the same data formats.

Cloud platforms also won’t interoperate because their VMs don’t interact in a standard way with different network and storage archi­tectures, APIs, network connections, databases, and other elements. VM translation is an important issue to enable the preservation of security policy, network policy, and identity across clouds. Without standardization, moving a workload from one cloud platform to another requires creating a new VM on the second platform and then reinstalling the application, which can take considerable time and effort.