Page 854 - Cloud computing: From paradigm to operation
P. 854

5                                            Intercloud and interoperability


            5       Conventions

            In this Recommendation:
            The keywords ''is required to'' indicate a requirement which must be strictly followed and from which no
            deviation is permitted if conformance to this document is to be claimed.
            The keywords ''is recommended'' indicate a requirement which is recommended but which is not absolutely
            required. Thus, this requirement need not be present to claim conformance.


            6       Overview of inter-cloud trust management
            The  inter-cloud  computing  concept  is  based  on  the  relationship  (pattern)  among  multiple  cloud  service
            providers (CSPs). This pattern (peering, federation or intermediary) allows a CSP to interwork with one or
            more peer CSPs to assure intermediation and security of services provided by these CSPs. The trusted inter-
            cloud relationship among multiple CSPs relies on confidence between CSPs, or between  a cloud service
            customer  (CSC)  and  a  CSP,  as  one  of  them  has  to  delegate  physical  control  over  applications,  services,
            resources and data to the others. Therefore, trust management is a key point between CSPs, or between a
            CSC and a CSP in inter-cloud.

            The trust management of inter-cloud uses a two-dimensional (vertical and horizontal) model, where the
            vertical axis is based on the layers of the cloud computing reference architecture [ITU-T Y.3502], and the
            horizontal ones is based on the interconnection of CSPs based on the inter-cloud framework [ITU-T Y.3511].

            The inter-cloud trust management framework relies on components for managing isolation and security
            mechanisms,  which  handle  cross-layer  trust  and  establish  chain  of  trust,  respectively  [ITU-T Y.3514].
            According to particular needs, inter-cloud trust management may either be CSC-related or CSP-related.
            From a CSC-related perspective, major threats concern data security and privacy. Even assuming a trusted
            cloud service provider, a malicious administrator has sufficient permissions to affect unauthorised use and
            to  modify  customer  information.  Thus,  CSP  control  over  the  infrastructure  should  be  limited  to  avoid
            inspection  or  analysis  of  user  data  and  virtual  machine  (VM)  instances  without  explicit  CSC  consent.
            Moreover, security should be self-service, so that a CSC can exercise fine-grained control over protection of
            their cloud resources according to a given security service level agreement (SLA). This means the CSC should
            be  able  to  actively monitor  its  allocated  resources, while enabling  a  high  level  of  customizability of the
            architecture and its services.

            From  a  CSP-related  perspective,  major  threats  are  malicious  customers  issuing  attacks  against  the
            virtualization layer, to break isolation or perform denial-of-service (DoS). An attacker is a non-privileged
            malicious cloud tenant. Mitigation of such threats implies first isolation: tenants should not be authorized to
            monitor or modify VM state or data from other tenants. It also means a small trusted computing base: the
            number of vulnerabilities and failures in the platform is directly linked with the code size run at the highest
            privilege level and the set of primitives exported.
            Many different classes of mechanisms and architectures have been proposed to address this issue, where
            trust management and isolation are intimately linked. Representative isolation architectures include modular
            hypervisors partly controlled by the CSC or secure enclaves based on hardware security mechanisms. Side-
            channel attacks also introduce major isolation issues for inter-cloud systems. Although many cryptographic
            solutions protect user confidentiality in the cloud, it is not possible to perform efficient and fast enough
            arbitrary computations on the data while they are encrypted.
            In an inter-cloud environment, reputation is a key factor for selecting the CSP to transact with. There are
            different approaches to implement reputation-based trust management in inter-cloud environments, all of
            them are based on a uniform cloud SLA framework which could make CSP and CSC avoid confusion and have
            a common understanding about the cloud service quality commitment.









            846
   849   850   851   852   853   854   855   856   857   858   859