Research

Last week I started talking about data security in the cloud, and I referred back to our Data Security Lifecycle from back in 2007. Over the next couple of weeks I’m going to walk through the cycle and adapt the controls for cloud computing. After that, I will dig in deep on implementation options for each of the potential controls. I’m hoping this will give you a combination of practical advice you can implement today, along with a taste of potential options that may develop down the road. We do face a bit of the chicken and egg problem with this series, since some of the technical details of controls implementation won’t make sense without the cycle, but the cycle won’t make sense without the details of the controls. I decided to start with the cycle, and will pepper in specific examples where I can to help it make sense. Hopefully it will all come together at the end. In this post we’re going to cover the Create phase: Definition Create is defined as generation of new digital content, either structured or unstructured, or significant modification of existing content. In this phase we classify the information and determine appropriate rights. This phase consists of two steps – Classify and Assign Rights. Steps and Controls   < div class=”bodyTable”> Control Structured/Application Unstructured Classify Application Logic Tag/Labeling Tag/Labeling Assign Rights Label Security Enterprise DRM   Classify Classification at the time of creation is currently either a manual process (most unstructured data), or handled through application logic. Although the potential exists for automated tools to assist with classification, most cloud and non-cloud environments today classify manually for unstructured or directly-entered database data, while application data is automatically classified by business logic. Bear in mind that these are controls applied at the time of creation; additional controls such as access control and encryption are managed in the Store phase. There are two potential controls: Application Logic: Data is classified based on business logic in the application. For example, credit card numbers are classified as such based on on field definitions and program logic. Generally this logic is based on where data is entered, or via automated analysis (keyword or content analysis) Tagging/Labeling: The user manually applies tags or labels at the time of creation e.g., manually tagging via drop-down lists or open fields, manual keyword entry, suggestion-assisted tagging, and so on. Assign Rights This is the process of converting the classification into rights applied to the data. Not all data necessarily has rights applied, in which cases security is provided through additional controls during later phases of the cycle. (Technically rights are always applied, but in many cases they are so broad as to be effectively non-existent). These are rights that follow the data, as opposed to access controls or encryption which, although they protect the data, are decoupled from its creation. There are two potential technical controls here: Label Security: A feature of some database management systems and applications that adds a label to a data element, such as a database row, column, or table, or file metadata, classifying the content in that object. The DBMS or application can then implement access and logical controls based on the data label. Labels may be applied at the application layer, but only count as assigning rights if they also follow the data into storage. Enterprise Digital Rights Management (EDRM): Content is encrypted, and access and use rights are controlled by metadata embedded with the content. The EDRM market has been somewhat self-limiting due to the complexity of enterprise integration and assigning and managing rights. Cloud SPI Tier Implications Software as a Service (SaaS) Classification and rights assignment are completely controlled by the application logic implemented by your SaaS provider. Typically we see Application Logic, since that’s a fundamental feature of any application – SaaS or otherwise. When evaluating your SaaS provider you should ask how they classify sensitive information and then later apply security controls, or if all data is lumped together into a single monolithic database (or flat files) without additional labels or security controls to prevent leakage to administrators, attackers, or other SaaS customers. In some cases, various labeling technologies may be available. You will, again, need to work with your potential SaaS provider to determine if these labels are used only for searching/sorting data, or if they also assist in the application of security controls. Platform as a Service (PaaS) Implementation in a PaaS environment depends completely on the available APIs and development environment. As with internal applications, you will maintain responsibility for how classification and rights assignment are managed. When designing your PaaS-based application, identify potential labeling/classification APIs you can integrate into program logic. You will need to work with your PaaS provider to understand how they can implement security controls at both the application and storage layers – for example, it’s important to know if and how data is labeled in storage, and if this can be used to restrict access or usage (business logic). Infrastructure as a Service (IaaS) Classification and rights assignments depend completely on what is available from your IaaS provider. Here are some specific examples: Cloud-based database: Work with your provider to determine if data labels are available, and with what granularity. If they aren’t provided, you can still implement them as a manual addition (e.g., a row field or segregated tables), but understand that the DBMS will not be enforcing the rights automatically, and you will need to program management into your application. Cloud-based storage: Determine what metadata is available. Many cloud storage providers don’t modify files, so anything you define in an internal storage environment should work in the cloud. The limitation is that the cloud provider won’t be able to tie access or other security controls to the label, which is sometimes an option with document management systems. Enterprise DRM, for example, should work fine with any cloud storage provider. This should give you a good idea of how to manage classification and

Got an IM from Rich today: “nasty windows flaw out there – worst in a long time”. I looked over the Microsoft September Security Bulletin and what was posted this morning on their Security Research and Defense blog, and it was clear he is right. MS09-045 and MS09-046 are both “drive-by style” vulnerabilities. The attack vector is most likely malicious websites hosting specially-crafted JavaScript (MS09-045) or malicious use of the DHTML ActiveX control (MS09-046) to infect browsing users. Vulnerabilities that confuse the script engine can be tough to reverse-engineer from the update so it may take a while for attackers to discover and weaponize. We still might see a reliable exploit within 30 days, hence the “1” … The attack vector for both CVE’s addressed by MS09-047 is most likely again a malicious website but these vulnerabilities could also be exploited via media files attached to email. When a victim double-clicks the attachment and clicks “Open” on the dialog box, the media file could hit the vulnerable code. I started writing up an analysis of the remotely exploitable threats, which can completely hose your system, when it dawned on me that technical analysis in this case is irrelevant. I hate to get all “Uh, remote code execution is bad, mmmkay” as that is unhelpful, but I think in this case, simplicity is best. Patch your Vista and Windows machines now! If you need someone else to tell you “Yeah, you’re screwed, patch now”, there is a nice post on the MSRC blog you can check out. If there is not an exploit in the wild already, I am not as optimistic as the MS staff, and think we will probably see something by week’s end. Share: