Data protection measures¶
In addition to purely technical security measures we implement additional measures to provide a safe environment for your and your customers’ data based on the following standards and regulations:
Certified Information Security Management System (ISMS) based on ISO/IEC 27001,
Germany’s federal data protection law (BDSG), and
Responsible bodies¶
Information Security officers:
Data protection officer:
General¶
This document describes the security measures required by GDPR Art. 32 (1). Based on our ISO 27001 certification we operate a risk-based ISMS leading to extensive security measures and documentation. We don’t replicate the entire documentation here but only the gist.
As some of the requirements of the law depend on the specific (contractual) situation, we’d like to first outline the general situation that hosting in the Flying Circus puts you in:
- Order data processing
applies within the framework of the GDPR. We will implement a contract for data processing suitable to the GDPR with you.
- Duty to give information
People who are affected by processing of data have right to request information about what data about them is being stored as well as about the processes how this data is being kept secure.
- Periodical check for compliance to regulations:
The client has the responsibility to routinely check that applicable regulations are being conformed to by their contractor. Yearly checks appear to be sufficient even for highly sensitive data (e.g. medical records). The result needs to be documented in written form. The audit usually needs to be performed by a third party.
Confidentiality¶
Admission control¶
Measures for admission control ensure that unauthorized persons cannot physically access the data processing equipment.
The physical assets (servers, switches, hard drives, …) are located in EU data centers operated by third parties. The ownership of the physical equipment is with the Flying Circus, or, in special cases by our customers [1].
For each data center used by us we require the following measures:
video surveillance (outside facilities, in-doors and rack corridors)
two-factor security for granting access (e.g. personal password and transponder card) with logging
physical access must be documented
preferably 24 hour guard services with linked alarm system
separate physical security zones for general areas, data center infrastructure and customer-accessible areas
separately locked racks with the possibility to use custom locks and keys.
We maintain a separate List of data centers.
Physical access to data processing equipment may be performed only by the Flying Circus’ administrators. Administrators may delegate physical access to other persons (e.g., data center staff).
Low-security locations¶
The Flying Circus operates one low-security zone called development (used for the development of the Flying Circus infrastructure).
Those locations are currently not allowed to be used for customer data storage as they are not sufficiently protected.
Due to this limitations, the services in high-security locations are not allowed to depend on services provided in low-security locations.
Entry control¶
Measures for entry control ensure that unauthorized third parties can not make use of the data processing systems.
Machines managed within the Flying Circus can be accessed by a variety of ways for management purposes: SSH, web interfaces, and others. For those we employ a homogeneous scheme to identify and authorize users within the Flying Circus. Management access to systems must use encrypted communication channels.
Identification and authorization of customer applications, not managed by the Flying Circus infrastructure, are not covered by our security responsibility. Our customers are required to ensure the security of their applications themselves.
User identification must be performed using personal credentials, so that actions can be traced back to an individual originator. Thus, sharing one’s credentials with another person is prohibited. Credentials can be either a username and a cryptographic measure (e.g. a private/public key scheme) or a password, depending on the applicability.
Users with a Flying Circus account are required to manage their password securely: Passwords must not become compromised if a device is being accessed unauthorizedly (logically or physically). Things to consider are for example: Home directory on a notebook, keychain or password manager software, backups, USB sticks, smartphones. Strongly encrypted storage of passwords is permitted and even advised. For Flying Circus staff there is a separate guideline for handling secret authentication information.
All hardware machines have emergency root logins which may only be used by Flying Circus administrators if regular user authentication is not working correctly. Such uses are monitored and must be documented.
All privileged actions need to be securely logged. For machines based on our current (NixOS) platform, this is achieved via a local logging journal, which cannot be tampered with by normal users. Additionally systems logs are shipped to a central log server within the same site where the logs are analysed and monitored.
SSH logins must be performed using SSH keys. Password authentication is not allowed and prevented by the system configuration. Successful SSH logins to machines are logged, unsuccessful SSH login attempts are not [2]. Excessive unsuccessful SSH login attempts automatically cause a blocking via firewall rules.
Access control¶
Measures for access control protect against access by unauthorized personnel.
Customer-owned virtual machines may be accessed by all Flying Circus administrators implicitly. In projects additional staff (e.g. support staff) may get explicit access. Access by others (e.g., customer personnel, third parties) must be authorised by a client representative.
Users are centrally managed using https://my.flyingcircus.io. Users are automatically provisioned to all relevant systems, including proper removal of access rights.
Flying Circus implements a permission-based concept to separate application maintenance tasks from privileged administrative tasks: for example, customer software updates or database access versus OS updates or OS configuration.
Privileged administrative access is generally not granted to customers. In cases where another person who is not an administrator is needed to solve a problem, a shared session between an administrator and the other person must be established (e.g. with screen).
Technically, there are three access variants to perform privileged administrative operations:
Using a user account which has been granted the ‘login’ and ‘wheel’ permissions for a certain project. This requires the user to log into a regular account using his SSH key and additionally provide his password to access privileged operations.
Using a user account which is member of the global group of administrators (see List of administrators) which grants access to all machines within the Flying Circus infrastructure.
Emergency root logins (see above in Entry control).
Authorized and unauthorized access to privileged operations is logged and analysed on a central loghost within the same site. [3]
Flying Circus maintains a set of permissions which enable users to perform application maintenance and other semi-privileged tasks, e.g. access to service user accounts or database administration rights. Permissions are granted to individual users by the customer or upon customer request.
All permission assignments are traceable and explicitly documented: their effects are documented in the configuration code and their assignments are documented in the configuration database. A comprehensive list of users and their permissions may be produced automatically on request.
Group accounts are generally not allowed to perform privileged administrative operations to ensure traceability of actions.
Passwords for physical machines granting access to root accounts and IPMI controllers are stored as copies in a strongly encrypted password manager.
Separation¶
Measures for separation ensure that data that is collected for separate purposes must be processed separately.
To separate data from different customers the Flying Circus facilitates virtualization: both virtual machines (to separate execution context) and SAN (to separate storage) ensure that customers can only access data belonging to them. Within a single machine access to different files and processes is available using standard UNIX permissions.
Machines (both virtual and physical) live in a specific access ring (short: ring):
Ring 0 machines perform infrastructure tasks. Thus, they need to process data belonging to several customers. Only administrator access is allowed on such machines. Examples include VM hosts and storage servers.
Ring 1 machines process data for a specific customer and are accessible to users associated to that customer. Examples include customer VMs.
All resources that belong logically together (e.g., VMs, storage volumes) are bundled into resource groups. Resource groups share that same set of user accounts and permissions.
Pseudonymisation¶
Measures to ensure that personal data can no longer be attributed to a specific person without the use of additional information.
We delete data after the retention times required by tax or commercial law.
Access log files are being anonymised.
We delete customer data upon customer’s request.
Integrity and Authenticity¶
Transfer control¶
Measures for transfer control ensure that data that is being saved or transferred is protected against unauthorized reading, copying, modification, or deletion. It also requires that the points for intentional transfer are documented.
All private data transferred past the boundary of a machine must use an authenticated and encrypted communication channel (exceptions see below). Data paths where sensitive information may be transferred include:
Application data (e.g., database contents) is transferred from or to the customer using the standardised encrypted protocols, e.g., SCP/SFTP, https.
Persistent data is saved on storage servers. Storage traffic is not encrypted due to performance reasons. Storage servers are connected to application servers using a private network. Machines on which administrative privileges are granted to customers are not allowed to connect directly to the storage network (see also Network security).
Backups are transferred to backup servers at the same site using either an encrypted communication channel or the private storage network. Backup data may also be transferred to off-site backup servers to improve disaster recovery abilities. The transfer must be encrypted.
In addition to application data, a system can generate data at runtime that contains sensitive information, for example log files. Log files usually do not leave the machine on which they were generated, unless the customer operates a logging server. Log data may also be transferred to a log server on the same site operated by Flying Circus via an encrypted channel. Only Flying Circus administrators may have access to the central log server.
Input control¶
Measures for input control ensure that input, change, and deletion of data are documented showing at least who worked when on what data.
The security of data entry, change and deletion is generally part of the customer’s application. Customers must ensure that data entry, deletion and removal are handled appropriately according to their applicable data protection laws.
However, during maintenance work it may be necessary that administrators need to enter, change, or delete data records on a low technical level to ensure the continued operation of the overall system. This will only happen after having informed the affected customers and having documented this in our issue tracking system.
Managed log files are rotated by the Flying Circus infrastructure automatically with sensible retention times.
Changes in the Flying Circus user directory (e.g., SSH keys) can be performed by the customer themselves or through our support. If the change happens through our support then it must be documented beforehand and confirmed by the customer after the change has been performed.
Order control¶
Measures for order control ensure that data is only processed according to the orders of the client.
The Flying Circus ensures that all actions taken by system administrators are covered by a contract or order with the customers affected by the action. This can be due to broad maintenance contracts or due to specific support requests.
Individual change requests should have an associated ticket in the Flying Circus request tracking system. Other means of documentation to control changes are possible, e.g., explanatory commit messages in a version control system.
Specific actions performed will be reported to the customer if required.
Availability¶
Measures for availability ensure that data is not accidentally destroyed or lost.
The availability of resources depending on the data center facilities is delegated to the operator of the data center. The Flying Circus facilitates service level agreements to make expectations about availability explicit.
The selection of hardware is performed by the Flying Circus using professional equipment and vendors. The Flying Circus facilitates standard procedures for increased availability of single components (e.g., RAID storages, redundant power supplies, spare components).
Customer data is regularly backed up according to the Flying Circus’ backup schedule. Restoration of past states may be performed by administrators on request. Additionally, a disaster recovery plan details failure scenarios, our preventative and recovery measures.
Other measures¶
Our support process and incident response measures are documented at the support overview.
We have a process for emergency and crisis management including contingency plans for critical business processes (business continuity). See also Disaster recovery.
Footnotes