Cloud Security Alliance The Cloud Computing Threat Vector Jim Reavis, Executive Director September 2009

About the Cloud Security Alliance Global, not-for-profit organization Inclusive membership, supporting broad spectrum of subject matter expertise: cloud experts, security, legal, compliance, virtualization, and on and on… We believe Cloud Computing has a robust future, we want to make it better “To promote the use of best practices for providing security assurance within Cloud Computing, and provide education on the uses of Cloud Computing to help secure all other forms of computing.”

Getting Involved Individual Membership (free) Subject matter experts for research Interested in learning about the topic Administrative & organizational help Corporate Members Help fund outreach, events Participate in Solution Provider Advisory Council Affiliated Organizations (free) Joint projects in the community interest

Members Over 4,000 members Broad Geographical Distribution Active Working Groups Editorial Educational Outreach Architecture Governance, Risk Mgt, Compliance, Business Continuity Legal & E-Discovery Portability, Interoperability and Application Security Identity and Access Mgt, Encryption & Key Mgt Data Center Operations and Incident Response Information Lifecycle Management & Storage Virtualization and Technology Compartmentalization New Working Groups Healthcare Cloud Threat Analysis Government Financial Services 4

Project Roadmap April 2009: Security Guidance for Critical Areas of Focus for Cloud Computing – Version 1 October 2009: Security Guidance for Critical Areas of Focus for Cloud Computing – Version 2 October 2009: Top Ten Cloud Threats (monthly) November 2009: Provider & Customer Checklists December 2009: eHealth Guidance December 2009: Cloud Threat Whitepaper Global CSA Executive Summits Q1 2010 – Europe Q1 or Q2 2010 - US 5

What is Cloud Computing? Not “One Cloud”: Nuanced definition critical to understanding risks & mitigation Working definition: Cloud describes the evolutionary development of many existing technologies and approaches to computing that separates application and information resources from the underlying infrastructure and mechanisms used to deliver them. This separation of resources from infrastructure combined with a utility- like, elastic allocation model creates a compelling model for Internet scale computing. The working definition is a bit ‘wordy’, but this is the official definition 5 essential characteristics: on-demand self-service (customer can provision computing capabilities without vendor interaction), broad network access (standards-based and with varied client support), resource pooling (multi-tenant model, location independence), rapid elasticity (in some cases automatic scaling – up or down), and measured service (resource monitoring and metering). 3 delivery models: Infrastructure as a Service (IaaS) (vendor provides infrastructure, customer deploys arbitrary software, including operating systems and applications), Platform as a Service (PaaS) (customer deploys applications created using programming languages and tools provided by the vendor), Software as a Service (SaaS) (vendor provides and manages applications and customer has control over specific user application configuration settings only) 4 deployment models: Public (available to general public), Private (operated solely for a single organisation), Managed (shared among several organisations and managed by one of them or an external vendor), Hybrid (a composition of two or more clouds)

Defining the Cloud On demand usage of compute and storage 5 principal characteristics (abstraction, sharing, SOA, elasticity, consumption/allocation) 3 delivery models Infrastructure as a Service (IaaS) Platform as a Service (PaaS) Software as a Service (SaaS) 4 deployment models: Public, Private, Hybrid, Community 7

Infrastructure as a Service You “RFP” security in S-P-I Model SaaS Software as a Service You build security in PaaS Platform as a Service The security approach and role varies depending on the delivery model IaaS Infrastructure as a Service 8

Key Challenges We aren’t moving to the cloud.. We are reinventing within the cloud Confluence of technology and economic innovation Disrupting technology and business relationships Pressure on traditional organizational boundaries “Gold Rush” mentality, backing into 20 year platform choice Challenges traditional thinking How do we build standards? How do we create architectures? What is the ecosystem required to managed, operate, assess and audit cloud systems? 9

Lots of Governance Issues Cloud Provider going out of business Provider not achieving SLAs Provider having poor business continuity planning Data Centers in countries with unfriendly laws Proprietary lock-in with technology, data formats Mistakes made by internal IT security – several orders of magnitude more serious A number of threats arise when you implement a cloud computing model Some of these are not immediately obvious Data leakage due to breach (hacker attack), but also a result of exposure of data in transit, and disclosure due to lawsuit Data retention is a double-edged sword – need a policy for data destruction also If accounts are not managed correctly it may not be obvious that a ‘stale’ account has been compromised A DDoS attack has a financial impact as the metering aspect means that service cost will increase! Virtualisation means that compromising a single piece of infrastructure can expose many customers The trust issue comes up again, both in terms of the provider staff and the provider infrastructure and application security Increased use of ‘web 2.0’ applications and services introduces new risks – the law of unintended consequences! 10

Thinking about Threats Technology Unvetted innovations within the S-P-I stack Well known cloud architectures Business How cloud dynamism is leveraged by customers/providers E.g. provisioning, elasticity, load management Old threats reinvented: “must defend against the accumulation of all vulnerabilities ever recorded”, Dan Geer- ism Malware in the cloud, for the cloud Lots of blackbox testing A number of threats arise when you implement a cloud computing model Some of these are not immediately obvious Data leakage due to breach (hacker attack), but also a result of exposure of data in transit, and disclosure due to lawsuit Data retention is a double-edged sword – need a policy for data destruction also If accounts are not managed correctly it may not be obvious that a ‘stale’ account has been compromised A DDoS attack has a financial impact as the metering aspect means that service cost will increase! Virtualisation means that compromising a single piece of infrastructure can expose many customers The trust issue comes up again, both in terms of the provider staff and the provider infrastructure and application security Increased use of ‘web 2.0’ applications and services introduces new risks – the law of unintended consequences! 11

Evolving Threats 1/2 Unprotected APIs / Insecure Service Oriented Architecture Hypervisor Attacks L1/L2 Attacks (Cache Scraping) Trojaned AMI Images VMDK / VHD Repurposing Key Scraping Infrastructure DDoS A number of threats arise when you implement a cloud computing model Some of these are not immediately obvious Data leakage due to breach (hacker attack), but also a result of exposure of data in transit, and disclosure due to lawsuit Data retention is a double-edged sword – need a policy for data destruction also If accounts are not managed correctly it may not be obvious that a ‘stale’ account has been compromised A DDoS attack has a financial impact as the metering aspect means that service cost will increase! Virtualisation means that compromising a single piece of infrastructure can expose many customers The trust issue comes up again, both in terms of the provider staff and the provider infrastructure and application security Increased use of ‘web 2.0’ applications and services introduces new risks – the law of unintended consequences! 12

Evolving Threats 2/2 Web application (mgt interface!) Data leakage XSRF XSS SQL Injection Data leakage Poor account provisioning Cloud provider insider abuse Financial DDoS "Click Fraud” A number of threats arise when you implement a cloud computing model Some of these are not immediately obvious Data leakage due to breach (hacker attack), but also a result of exposure of data in transit, and disclosure due to lawsuit Data retention is a double-edged sword – need a policy for data destruction also If accounts are not managed correctly it may not be obvious that a ‘stale’ account has been compromised A DDoS attack has a financial impact as the metering aspect means that service cost will increase! Virtualisation means that compromising a single piece of infrastructure can expose many customers The trust issue comes up again, both in terms of the provider staff and the provider infrastructure and application security Increased use of ‘web 2.0’ applications and services introduces new risks – the law of unintended consequences! 13

CSA Guidance Domains Understand Cloud Architecture Governing in the Cloud Governance & Risk Mgt Legal Electronic Discovery Compliance & Audit Information Lifecycle Mgt Portability & Interoperability Operating in the Cloud Traditional, BCM, DR Data Center Operations Incident Response Application Security Encryption & Key Mgt Identity & Access Mgt Storage Virtualisation To effectively address the risks posed by all possible cloud computing models, countermeasures need to be implemented across a range of areas The Cloud Security Alliance have developed a framework for the security controls and countermeasures The framework is outlined on the slide shown, and can be broken down into two main streams – Governance, and Operations Overall requirement to understand the architecture 14

Governance & ERM A portion of cloud cost savings must be invested into provider scrutiny Third party transparency of cloud provider Financial viability of cloud provider. Alignment of key performance indicators Increased frequency of 3rd party risk assessments

Legal Plan for both an expected and unexpected termination of the relationship and an orderly return of your assets. Find conflicts between the laws the cloud provider must comply with and those governing the cloud customer Gain a clear expectation of the cloud provider’s response to legal requests for information. Secondary uses of data Cross-border data transfers

Electronic Discovery Cloud Computing challenges the presumption that organizations have control over the data they are legally responsible for. Cloud providers must assure their information security systems are capable to preserve data as authentic and reliable. Metadata, logfiles, etc. Mutual understanding of roles and responsibilities: litigation hold, discovery searches, expert testimony, etc.

Compliance & Audit Classify data and systems to understand compliance requirements Understand data locations, copies Maintain a right to audit on demand Need uniformity in comprehensive certification scoping to beef up SAS 70 II, ISO 2700X

Information Lifecycle Mgt Understand the logical segregation of information and protective controls implemented Understand the privacy restrictions inherent in data entrusted to your company, how it impacts legality of using cloud provider. Data retention assurance easy, data destruction may be very difficult. Recovering true cost of a breach: penalties vs risk transference

Portability & Interoperability Understand and implement layers of abstraction For Software as a Service (SaaS), perform regular data extractions and backups to a usable format For Infrastructure as a Service (IaaS), deploy applications in runtime in a way that is abstracted from the machine image. For Platform as a Service (PaaS), careful application development techniques and thoughtful architecture should be followed to minimize potential lock-in for the customer. “loose coupling” using SOA principles Understand who the competitors are to your cloud providers and what their capabilities are to assist in migration. Advocate open standards.

Traditional, BCM/DR Greatest concern is insider threat Cloud providers should adopt as a security baseline the most stringent requirements of any customer. Compartmentalization of job duties and limit knowledge of customers. Onsite inspections of cloud provider facilities whenever possible. Inspect cloud provider disaster recovery and business continuity plans. Identify physical interdependencies in provider infrastructure.

Data Center Operations Compartmentalization of systems, networks, management, provisioning and personnel. Know cloud provider’s other clients to assess their impact on you Understand how resource sharing occurs within your cloud provider to understand impact during your business fluctuations. For IaaS and PaaS, the cloud provider’s patch management policies and procedures have significant impact Cloud provider’s technology architecture may use new and unproven methods for failover. Customer’s own BCP plans should address impacts and limitations of Cloud computing. Test cloud provider’s customer service function regularly to determine their level of mastery in supporting the services.

Incident Response Any data classified as private for the purpose of data breach regulations should always be encrypted to reduce the consequences of a breach incident. Cloud providers need application layer logging frameworks to provide granular narrowing of incidents to a specific customer. Cloud providers should construct a registry of application owners by application interface (URL, SOA service, etc.). Cloud providers and customers need defined collaboration for incident response.

Application Security Importance of secure software development lifecycle maganified IaaS, PaaS and SaaS create differing trust boundaries for the software development lifecycle, which must be accounted for during the development, testing and production deployment of applications. For IaaS, need trusted virtual machine images. Apply best practices available to harden DMZ host systems to virtual machines. Securing inter-host communications must be the rule, there can be no assumption of a secure channel between hosts Understand how malicious actors are likely to adapt their attack techniques to cloud platforms

Encryption & Key Mgt From a risk management perspective, unencrypted data existent in the cloud may be considered “lost” by the customer. Application providers who are not controlling backend systems should assure that data is encrypted when being stored on the backend. Use encryption to separate data holding from data usage. Segregate the key management from the cloud provider hosting the data, creating a chain of separation. When stipulating standard encryption in contract language

Identity & Access Mgt Must have a robust federated identity management architecture and strategy internal to the organization. Insist upon standards enabling federation: primarily SAML, WS-Federation and Liberty ID-FF federation Validate that cloud provider either support strong authentication natively or via delegation and support robust password policies that meet and exceed internal policies. Understand that the current state of granular application authorization on the part of cloud providers is non-existent or proprietary. Consider implementing Single Sign-on (SSO) for internal applications, and leveraging this architecture for cloud applications. Using cloud-based “Identity as a Service” providers may be a useful tool for abstracting and managing complexities such as differing versions of SAML, etc.

Storage Understand the storage architecture and abstraction layers to verify that the storage subsystem does not span domain trust boundaries. Ascertain if knowing storage geographical location is possible. Understand the cloud provider’s data search capabilities. Understand cloud provider storage retirement processes. Understand circumstances under which storage can be seized by a third party or government entity. Understand how encryption is managed on multi-tenant storage. Can the cloud provider support long term archiving, will the data be available several years later?

Virtualization Virtualized operating systems should be augmented by third party security technology. The simplicity of invoking new machine instances from a VM platform creates a risk that insecure machine images can be created. Secure by default configuration needs to be assured by following or exceeding available industry baselines. Virtualization also contains many security advantages such as creating isolated environments and better defined memory space, which can minimize application instability and simplify recovery. Need granular monitoring of traffic crossing VM backplanes Provisioning, administrative access and control of virtualized operating systems is crucial

Lots of work to do New cloud providers Easy to bypass IT Need agile view of systems Need executive involvement Need standards Need to learn from past mistakes Do visit the website Do join the LinkedIn Groups – you will receive regular email updates

Contact www.cloudsecurityalliance.org info@cloudsecurityalliance.org Twitter: @cloudsa, #csaguide LinkedIn: www.linkedin.com/groups?gid=1864210 Do visit the website Do join the LinkedIn Groups – you will receive regular email updates

Thank You!