Crosswalk pair
FedRAMP 20x and ISO/IEC 27001, control by control
25 canonical controls in Keel’s library satisfy clauses of both FedRAMP 20x and ISO/IEC 27001. Implement each once, attach the evidence once, and it counts toward each standard. The overlap is the work you don’t repeat.
The overlap
What the two libraries have in common
Every figure here counts canonical controls in Keel’s library, not clauses of either standard. Each standard’s own authored count is on its framework page.
25
Controls that satisfy both
Canonical controls that crosswalk to at least one clause of each.
27
In Keel’s library for FedRAMP 20x
93% of them also map to ISO/IEC 27001.
84
In Keel’s library for ISO/IEC 27001
30% of them also map to FedRAMP 20x.
103
Evidence artifacts expected
Across the shared controls, from Keel’s evidence guidance. Gathered once.
-
FedRAMP 20x 93%
25 controls of 27 in Keel’s library for FedRAMP 20x also map to ISO/IEC 27001.
-
ISO/IEC 27001 30%
25 controls of 84 in Keel’s library for ISO/IEC 27001 also map to FedRAMP 20x.
The mapping
Controls that satisfy both
Each row is one control in Keel’s library and the clauses it answers on each side. Do the work once; both columns are then evidenced by the same artifacts.
| Canonical control | FedRAMP 20x clauses | ISO/IEC 27001 clauses |
|---|---|---|
| Governance & Risk | ||
| Continual improvement Improvement of the management system is run as an activity with a record, not held as an intention. Opportunities are captured from everywhere they arise - audit findings, the results of measurement and evaluation, decisions out of management reviews, incidents and near misses, and suggestions from the people actually doing the work - and held in one place instead of in the meeting each came out of. Each is evaluated and either taken forward with an owner and a date or closed with the reason it was not, so a rejected idea is a decision rather than a silence. Once an improvement is made, its effect on the suitability, adequacy and effectiveness of the system is checked, so improvement is something that can be shown to have happened rather than asserted at the next audit. The routine execution of the work counts as a source in its own right: what the operational processes, procedures and activities themselves show while they are being run - the step that is always skipped, the check that never fires, the manual workaround everybody has quietly adopted - is captured on the same terms as a finding from an audit, because the people running a process daily see more of it than any evaluation does. Where the thing being improved is an AI system, the improvement activity is built into the system’s own update cycle and is measurable: an update carries the improvement it is meant to deliver and the measure that will show whether it did, checked afterwards rather than asserted, and interested parties - the people who operate the system, the people it is used on, and those who represent them - are engaged regularly as part of that cycle rather than consulted once at launch. The opportunities are not confined to the system: what the organization delivers is in scope too. Improving products and services to meet known requirements and to address needs and expectations that are coming rather than current, correcting, preventing or reducing undesired effects, and improving how the system itself performs are all determined and selected against as improvement opportunities, rather than run as three unrelated programs. And the standing question - what in the system’s suitability, adequacy and effectiveness should be improved next - is answered from evidence rather than appetite: the results of analysis and evaluation and the outputs of management review are considered together to decide whether there is a need or an opportunity that has to be addressed. | KSI-SVC-EIS | 10.1 |
| Leadership commitment & accountability Top management is accountable for whether the management system works, and the accountability is exercised rather than asserted. It approves the policy and the objectives and satisfies itself that they fit the direction the organization is actually going in; it requires the system’s requirements to be built into how the business already runs rather than bolted alongside it; it makes sure the resources the system needs are available; it tells the organization why conforming to the system matters, in its own voice; it directs and supports the people whose work makes the system effective; it promotes improvement; and it backs other managers in exercising leadership over the parts of the system that are theirs. What it decided, when, and on what basis is recorded, so the commitment is evidenced by acts rather than by a signature on a policy. Leadership is accountable for cybersecurity risk itself and not only for the system that manages it, and it fosters the culture that has to go with that: risk-aware, ethical, and expected to keep improving rather than to hold a standard once reached. It also requires cybersecurity risk management activities and their outcomes to be carried inside the organization’s enterprise risk management process - the same register, the same reporting line, the same committee that hears the other risks - rather than in a security-only record that never reaches the people who allocate capital against it. It promotes two habits by name rather than by implication: the process approach, meaning the work is understood, run and improved as connected processes with defined inputs, outputs and owners; and risk-based thinking, meaning what could go wrong is considered while the work is being planned rather than after it has gone wrong. And it satisfies itself that the system achieves the results it was set up to achieve - the intended outcomes, checked - so leadership answers for the system’s effect and not only for its existence. | KSI-PIY-RES | 5.1 |
| Resources for the management system The resources the management system needs in order to be established, run, kept running and improved are determined and provided, not assumed: the people and the time they are actually given rather than the time the plan says they have, the tools and technology, the information, and the budget. The determination distinguishes what the organization can meet from its own capability from what it has to obtain from outside, and it is written down so a shortfall is visible as a shortfall. It is revisited when the system’s scope, its workload or the organization changes, so a system that has grown is not still resourced for the size it was when it started. Security and privacy are budgeted as a discrete line rather than absorbed into a general technology allocation. The high-level security and privacy requirements for a system or a service are determined while the business process it serves is being planned, rather than after the design is fixed; what it will cost to protect it is then determined, documented and allocated as part of the organization’s capital planning and investment process; and that amount appears as a discrete line item in the programming and budgeting record - so an underfunded control is a visible decision rather than an unexplained gap. Adequacy is judged against the risk strategy rather than against last year’s allocation: what is provided is set commensurate with the risks the organization has said it will manage, the roles it has assigned and the policies it has issued - and where it is not, the shortfall is recorded against the part of the strategy it fails to fund. People are determined as their own class of resource rather than counted inside a budget line: the persons necessary for the system to be implemented effectively, and for its processes to be operated and controlled, are identified from the work that has to be done and are then actually provided - so a process with nobody assigned to run it is visible before it fails rather than after. Where the management system depends on data and on computing capacity - not only on people, tools and money - those are determined and provided as resource classes in their own right, so a system planned without the data it needs, or without the compute to run what it plans, is a visible shortfall rather than a later discovery. | KSI-PIY-RIS | 7.1 |
| Access Control | ||
| Access control policy Rules for granting, reviewing, and revoking access to systems and data based on business need and least privilege; anyone who works with sensitive data, or in a place from which it can be reached, is individually authorized for that work or supervised while doing it; and a documented emergency route exists to obtain the data when the normal access path is unavailable, with every use of that route recorded and reviewed afterwards. The rules also settle the opposite question: which actions, if any, a person may take on a system without identifying or authenticating themselves at all. Those actions are identified rather than left as whatever the system happens to permit, they are limited to what the organization’s business actually requires, and each one is documented in the system’s security plan together with the reasoning that justifies it, so an unauthenticated path is a decision somebody made and can be asked about. The rules are enforced by access control lists set on the data itself, not only by what an application chooses to show: permissions on local and remote file systems, on databases and inside applications are configured to the holder’s need to know, so information a role has no business reading is unreachable rather than merely unadvertised. What an authorized user may DO is limited on the same terms as what they may read: the types of transaction and function each role is permitted to execute are decided in advance and enforced by the system, so holding access to an application does not carry the right to run every operation inside it, and an action outside the permitted set is refused rather than merely unadvertised in the interface. Enforcement is centralized wherever the systems support it - access decisions for enterprise assets are made by a single directory service or single sign-on provider rather than by each system keeping its own list - and the systems that make those decisions are themselves known: an inventory of the organization’s authentication and authorization systems is maintained, including those run by a service provider on its behalf, and reviewed on a defined cadence. | KSI-IAM-ELP | A.5.15 |
| Multi-factor authentication Documented procedures verify that a person or system seeking access to sensitive data is the one it claims to be, on every path by which that data can be reached - and multi-factor authentication is the enforced mechanism for remote access, administrative access, and access to sensitive systems and data. The multi-factor mechanism itself is configured so it cannot be bypassed, so the factors it uses are genuinely independent of one another - one factor’s success granting no knowledge of and no route around another - and so access is refused unless every factor required has succeeded. The requirement is not confined to the accounts that carry privilege: every account is covered, privileged and non-privileged alike, because an ordinary account is the usual way into a privileged one. The mechanisms chosen are resistant to replay, so an authentication captured on the wire or lifted from a log cannot be presented again to gain access. Authentication is not a single event at the start of a session either: the person is required to authenticate again when the organization’s defined circumstances arise - a change of role or of the authenticators themselves, an escalation to privilege, a session that has run beyond its defined life, or a request to perform an action the organization has designated as requiring fresh proof. Where a regime requires the factors to be PHISHING-RESISTANT rather than merely replay-resistant, the organization uses factors that resist verifier impersonation, such as a hardware security key or a platform authenticator bound to the origin by public-key cryptography. A one-time code or a push approval does not count toward that requirement: both resist replay and neither survives an attacker relaying the exchange in real time. | KSI-IAM-APM | A.8.5 |
| Password & credential management Rules for the authentication credentials themselves: passwords are unique per account and meet a defined strength standard, a new or changed password is screened against a list of commonly used, expected and compromised passwords and refused if it appears there, a credential issued for first use must be replaced immediately, reuse of previous passwords is refused, changes follow a defined procedure, repeated failed authentication attempts lock the account for a defined period, and passwords, keys and other authentication secrets are stored and transmitted only in protected form. The rules are a documented policy with supporting procedures covering identification and authentication as a whole - who and what must be identified, to what assurance, and by which mechanisms - issued to the people and roles it applies to, owned by a named role, and reviewed and updated on a defined cadence and after a change that affects it. Repeated failure is handled to a stated rule rather than to a default: a limit is set on consecutive invalid authentication attempts within a defined time window, and when that limit is exceeded the system responds automatically in a way the organization has chosen and configured - locking the account for a defined period, delaying further attempts by a defined algorithm, or notifying a defined role - rather than continuing to accept attempts. What is displayed back to the person authenticating is obscured while they do it, so a credential cannot be read off the screen or recovered from the feedback the system gives. The protection an authenticator receives is proportionate to the sensitivity of what its use unlocks, so a credential granting access to the most sensitive information is stored, transmitted, issued and revoked under stronger handling than one that opens a low-impact system. Where a regime the organization is subject to sets an identity, authenticator or federation assurance LEVEL rather than leaving authenticator strength to judgement, the organization meets the level that regime names, at the grade it names for the particular tier of assurance the organization has committed to, since one regime can set a different level for each tier. Authentication assertions that pass through a third party, such as the browser, are encrypted in transit through it rather than only over the connection. | KSI-IAM-APM, KSI-IAM-SNU, KSI-SVC-ASM | A.5.17 |
| Privileged access & administrative tooling Access that can change a system rather than merely use it is treated as a separate class from ordinary access, and so are the tools that carry that power. Privileged rights are allocated through a documented process on an event-by-event basis rather than standing by default: they are requested against a stated need, approved by somebody other than the requester, granted for a defined period and withdrawn when it ends. They are held on an identity distinct from the person’s everyday account, so routine work does not run with administrative rights, and they require strong authentication including a second factor. Who holds what privilege is reviewed on a cadence and immediately on a role change or departure. Privileged sessions are logged in a way the privileged user cannot alter, and the log is reviewed. Emergency or break-glass access exists as a documented route, and every use is recorded and reviewed afterwards. Separately, UTILITY PROGRAMS capable of overriding system or application controls - system tools, database and configuration editors, debuggers, packet capture, recovery and imaging utilities - are inventoried, removed or disabled where they are not needed, restricted to a named authorized few, kept apart from application software, protected against being reintroduced by a user, and logged when used; anyone holding ordinary application access does not get them by default. Which personnel and roles may hold a privileged account at all is a decided and recorded list rather than an accumulation, and access to the system’s security functions and to security-relevant information - the audit configuration, the access rules, the cryptographic settings, the security tooling itself - is authorized explicitly per role rather than arriving as a side effect of an administrative grant. The periodic review asks whether the privilege is still needed for the work the holder actually does now, and reassigns or removes it where it is not. The technical controls match the rules: a user without privileges is prevented from executing a privileged function rather than only forbidden from doing so, and every execution of a privileged function is written to the log. Where privileged commands or access to security-relevant information are exercised over a remote connection, that is permitted only for needs the organization has documented and only in a form that can be audited afterwards. The separation extends to the machine and not only to the account: administrative work is performed from computing resources dedicated to it and separated physically or logically from the estate used for everyday work, segmented away from the organization’s primary network and without general internet access, so a browser or a mail client is never one process away from the session that administers the environment. ACCOUNTS THAT ARE NOT PEOPLE are governed as their own class rather than as a quiet exception to the rules above: every account used by a system, a service or an application is inventoried with the function it exists for and a named human owner, interactive login by a person is prevented on it unless a specific need is documented and authorized, its credentials are held in a managed secret store rather than in a file, a script, a scheduler or somebody’s memory, they are changed on a defined cadence and whenever a person who could have known them leaves, and they are neither shared between accounts nor reused from one environment to the next. | KSI-IAM-ELP, KSI-IAM-JIT, KSI-IAM-SUS, KSI-MLA-ALA | A.8.2, A.8.18 |
| User provisioning & deprovisioning Joiner/mover/leaver process to grant, change, and promptly remove access across systems, in which every person is issued an account of their own carrying a unique name or number, so an action in a log traces back to one named individual rather than to a shared or generic login. Each person’s right of access is recorded when it is established and reviewed on a schedule thereafter, as well as granted and changed - so what someone holds is a documented position that has been looked at again, not the accumulated residue of past requests - and what may be granted follows the organization’s access authorization rules rather than the judgment of whoever processes the request. Identity is managed as a lifecycle in its own right and not only as the access hung off it: an identity is created only after the person or the thing behind it has been verified to a stated standard, is linked to a single accountable human owner, and is disabled and then retired on a defined path rather than deleted where the record has to survive for an audit trail. A shared or generic identity exists only where there is a reason it cannot be individual, and then it is authorized, given an owner who answers for what is done with it, and reviewed. Identities issued to services, applications, devices and automation are registered on the same terms as human ones, with an owner, a purpose and a review date, because an unowned machine identity outlives every person who knew what it was for. Dormant identities are detected and removed rather than waiting for a leaver process that was never triggered. The lifecycle is run by automated mechanisms wherever the systems allow it: accounts are provisioned and deprovisioned from an authoritative source of record, and each act of creating, modifying, enabling, disabling or removing an account generates an audit record automatically rather than depending on the administrator to note it. An account issued for a temporary or emergency purpose carries an expiry from the moment it is created and is disabled or removed automatically when that expiry passes, so a route opened for one situation does not stay open after it. An account is disabled within a defined period when it has expired, when it is no longer associated with any individual, when it is in violation of the organization’s policy, or when it has been inactive beyond a defined period - and where an individual is found to pose a significant risk, within a defined period of that discovery rather than at the next scheduled review. Identifiers are managed as an object in their own right: an identifier is authorized before it is assigned, is selected to a defined convention, is never reused for a different person or entity, and is issued on the same terms whether it names an individual, a group, a role, a service or a device - and where the organization needs to distinguish one class of person from another, such as an employee from a contractor or a vendor, the identifier or its record carries that status rather than leaving it to be inferred. Identity proofing is performed to the assurance level the access warrants: the applicant is resolved to a single unique individual, is required to present identity evidence to whoever registers them, and that evidence is validated and verified by methods the organization has defined rather than accepted on sight, with an address of record confirmed through an out-of-band channel where the assurance level calls for it. All of it is answerable from one INVENTORY OF ACCOUNTS rather than from each system in turn: every account the organization manages is listed - ordinary user, administrator and service alike - with the person or function behind it, the account name, the dates it starts and stops, the department or owner it belongs to and the privilege it carries, and the list is validated against what is actually active on a defined recurring schedule, so an account nobody can account for is found by the review rather than by an incident. Accounts are managed centrally through a directory or identity service wherever a system can be brought into one, because an account that lives only inside an application is the one a leaver process misses. Where a regime requires the identifier itself to carry the STATUS of the person holding it, such as contractor, vendor or foreign national, the account identifier records that status rather than leaving it to a list kept somewhere else, so the status travels with every use of the account. | KSI-IAM-AAM, KSI-IAM-SUS | A.5.16, A.5.18, A.8.3 |
| Data Protection & Privacy | ||
| Data retention & secure disposal Data is retained per policy and securely destroyed when no longer needed. Retention periods are set against the purpose the data was collected for and any legal or contractual obligation to keep it, recorded per category of data rather than left to whoever is looking at the record, and enforced when they run out - data goes because its period ended, not because somebody finally objected to keeping it. Destruction leaves it unrecoverable rather than merely removed from an index, and what was destroyed, when, by what method and on whose authority is recorded. The hardware and media that held it reach a defined final disposition at end of life, by a route the organization has decided in advance rather than by whatever happens to the box; and any media that stays in service is cleared of that data before it is reused, reassigned, or passed to anyone else. Disposal is not confined to data and media: the documentation, the tools and the system components the organization has defined as needing it are disposed of by techniques and methods it has approved in advance - so a decommissioned appliance, a retired build server, a set of network diagrams or a licensed utility leaves the organization by a route somebody chose, and the route is recorded on the same terms as a data destruction. A retention period has two ends and both are stated: the minimum the organization must keep the data for, and the maximum beyond which it may not be kept - so retention is bounded in the direction of keeping too long as well as of destroying too early. | KSI-SVC-RUD | A.7.14, A.8.10 |
| Encryption in transit & at rest Strong cryptography protects sensitive data in transit over public networks and at rest in storage. The mechanisms are chosen to do two things and are judged against both: prevent unauthorized disclosure of the information, and prevent or detect unauthorized change to it - in transit, so a message altered between sender and receiver is caught rather than delivered, and at rest, so a stored record cannot be modified undetectably by somebody with access to the storage but not to the key. Which information is protected at rest, and on which system components, is decided and recorded rather than left to whatever the platform encrypts by default. The scope named explicitly reaches the end-user device as well as the server: data held on laptops, desktops and other end-user devices that carry sensitive information is encrypted at the device or volume level, so a device that leaves the building is an object somebody lost rather than a disclosure. And data in transit is encrypted wherever it is sensitive, not only where it crosses a public network - a session between two internal systems is protected on the same terms when what it carries warrants it. Where a law, a regulation or a contract requires the cryptography to be VALIDATED rather than merely strong, the organization uses a cryptographic module that carries the validation that instrument names, and it confirms that validation against the specific module, version and operating mode actually deployed rather than inferring it from the product’s name - because a validated module run outside the configuration it was validated in is not a validated module, and the certificate that proves the point is held as evidence rather than assumed to exist. | KSI-SVC-SIN, KSI-SVC-VCM, KSI-SVC-VRI | A.8.24 |
| Infrastructure & Operations | ||
| Asset inventory An inventory of hardware, software, and information assets with assigned owners, in which the movement of equipment and removable media into, out of, and within the organization’s premises is recorded against the person responsible for it. The record is maintained by the acts that change it rather than by a periodic sweep: installing a component, removing one, or updating a system updates the inventory as part of that work, and the estate is scanned for hardware, software and firmware that is present but not authorized, with the response to an unauthorized component - disable its network access, isolate it, remove it, notify a defined role - decided in advance. The inventory answers WHERE INFORMATION IS as well as what exists: for the categories of information the organization has defined as sensitive, it records where that information is processed and stored, which system components host it, and which users and roles can reach it, and it is updated when any of those change - with automated tools used to locate that information across components and to confirm the required protections are actually in place there, rather than relying on what a system was designed to hold. Each asset also carries its support status: the date vendor support ends is recorded, and a component that reaches end of support is replaced, or is covered by an alternative source of continued support that the organization has arranged and recorded, rather than left in service because it still runs. The inventory is kept current by discovery as well as by the acts that change it: an active discovery tool interrogates the network on a defined frequency, a passive tool identifies assets from the traffic they generate, the DHCP and address-management logs are read on a defined cadence so an address issued to something nobody registered surfaces, and automated software-inventory tooling documents what is installed across the estate rather than relying on a manual return. Support status decides authorization rather than merely being recorded: only software the vendor still supports is designated authorized in the inventory, and software that is unsupported and carries no documented exception setting out its mitigating controls and the residual risk somebody accepted is designated unauthorized, so the process that removes unauthorized software picks it up. What the register holds about those categories of information is itself a DATA INVENTORY: for each data type the organization has designated, the record carries the metadata that makes it usable - what the data is, who owns it, how it is classified, where it came from, what it is retained for and for how long - so the question of what data exists is answered from the register rather than from the systems one at a time. And the register tracks lifecycle STATE as well as existence: each system, device, piece of software, service and data set carries where it has reached in its life - requested, acquired, deployed, in service, superseded, withdrawn, disposed of - with the acts that move it between those states recorded against it, so an asset is managed from acquisition through operation to disposal rather than entered once and forgotten. | KSI-PIY-GIV | A.5.9 |
| Change management Changes to systems and software are requested, reviewed, tested, approved, and tracked. Each request is explicitly approved OR DISAPPROVED by a role authorized to decide it, rather than proceeding because nobody objected, and the decision, the reason behind it and the change itself are written to a durable record - so the log shows what was refused as well as what shipped, and a change that went in without a decision is visible as one. A proposed change is analyzed for what it would do to security and privacy BEFORE it is approved, so the decision is taken with that answer in hand rather than reported afterwards; where the change is significant, the analysis is recorded and the roles who own the affected controls see it. Change control is a body rather than a queue: security and privacy representatives are members of the group that approves changes, so the question is asked in the room instead of by exception. A change is tested, validated and documented before it is finalized and moved into the running system, in an environment that resembles the one it is going to. After it lands, the controls it touched are verified to be implemented correctly, still operating as intended, and still producing the result the organization’s security and privacy requirements call for - verified, not assumed from the fact that the deployment succeeded. Who may make a change at all is restricted: the physical and logical access needed to alter a system is defined, documented, approved and enforced, and it is a narrower set than the people who may use the system. EXCEPTIONS are managed on the same terms as changes: a departure from a standard, a baseline or a policy is requested, assessed for the risk it creates, approved by a role authorized to accept that risk, recorded with a scope and an expiry, and tracked to closure or to a deliberate renewal - so an exception is a change somebody decided rather than a state nobody revisits. | KSI-CMT-LMC, KSI-CMT-RVP, KSI-CMT-VTD, KSI-SVC-PRR | A.8.32 |
| Logging & monitoring Security-relevant events - including successful and failed log-in attempts - are logged, protected, retained, and reviewed for anomalies, and the discrepancies that review finds are reported to the people who act on them. The review runs on a defined cadence and covers the records of system activity as a set - the audit logs, the reports of who accessed what, and the record of security incidents - rather than the log stream alone. For those records to be correlated into one sequence of events, the systems producing them agree on the time: every in-scope system synchronizes its clock to a single approved reference source, the source and the tolerance the organization will accept are specified rather than left to defaults, and timestamps are recorded in an unambiguous form so a reader does not have to infer a time zone. Synchronization is monitored in its own right - a system that drifts beyond tolerance or loses its source raises an alert, because a clock that is wrong makes an investigation reach the wrong conclusion rather than no conclusion - and where equipment cannot be synchronized, its offset is known and recorded so its records can still be placed. Timestamps are generated from the system’s own clock, expressed in Coordinated Universal Time or a recorded offset from it, and cut to a granularity the organization has stated rather than to whatever the platform defaults to. All of this rests on a documented audit and accountability policy with supporting procedures, aligned with the laws and obligations that apply to the organization, issued to the roles it binds, owned by a named role, and reviewed on a defined cadence. WHAT A RECORD CONTAINS is specified rather than accepted: every audit record establishes what type of event occurred, when it occurred, where it occurred, the source it came from, the outcome - success or failure - and the identity of any individual, subject or object associated with it, plus whatever further fields the organization has decided it needs to reconstruct an event afterwards. Because every person holds an account of their own, the identity a record carries resolves to one named individual rather than to a shared or generic login, so an action can be traced to whoever actually took it and that person can be held accountable for it - which is the whole reason the identity field is mandatory rather than useful. WHICH EVENTS ARE LOGGED is decided and then kept under review rather than configured once: the set of event types selected for logging is agreed with the roles who investigate, is reviewed on a defined cadence and again after an incident that showed the set was wrong, and is updated as a result - so the log answers the questions being asked now instead of the ones somebody anticipated at build. Storage is sized for that: enough capacity is allocated to hold the volume produced for the retention period the organization has set, and records are retained for that period specifically so an investigation after the fact is possible and so regulatory and internal obligations are met, rather than for as long as the disk happens to last. When the logging process itself fails - the pipeline stops, the store fills, a source goes silent - a defined role is alerted within a defined time and the organization takes the response it decided on in advance, because a logging failure is the one failure the logs cannot tell you about. REVIEW AND ANALYSIS are supported by machinery rather than by reading. Automated mechanisms integrate the review, analysis and reporting of audit records into a single process, and records drawn from separate repositories are correlated so the organization sees one organization-wide picture of activity instead of several partial ones. A reduction and reporting capability supports on-demand review, analysis and reporting and the investigation of an incident, and it does so without altering the original records or their ordering; it lets an analyst filter, sort and search records by the criteria the organization has defined, so events of interest surface in time to matter. THE RECORDS THEMSELVES ARE PROTECTED as an asset. Audit information and the logging tools that produce it are protected from unauthorized access, modification and deletion, a defined role is alerted when evidence of tampering is detected, and the ability to manage the logging function - what is collected, what is retained, what is deleted - is restricted to a named subset of privileged users rather than being available to every administrator whose activity it records. MONITORING runs on top of the record. The organization monitors its systems to detect attack and indicators of potential attack, unauthorized local, network and remote connections, and use that is outside what it has authorized; it identifies that use against defined criteria for what unusual looks like. Inbound and outbound communications traffic is watched for those conditions specifically, because exfiltration and command traffic look ordinary unless somebody has said what ordinary is. Automated tools and mechanisms support analysis close to real time rather than at the next review, and when the system produces an indication of compromise or potential compromise a defined role is alerted. What monitoring finds is reported to the people who act on it, at the frequency the organization has set. WHICH SOURCES ARE COLLECTED is decided rather than left to whatever a platform emits by default. Access to information the organization has classified as sensitive is logged, including modification and disposal and not only reading. DNS queries, URL requests and command-line activity are collected where the asset supports it, because those three are what an investigation reconstructs an intrusion from, and network traffic flow records are collected from the network devices so that movement between systems can be reviewed and alerted on. Logs from the service providers the organization depends on are collected too, so authentication, user-management and data-lifecycle events that happen outside its own estate sit inside the same record. Collection and retention are centralized so far as the estate allows, into a platform that correlates sources rather than storing them side by side, and security event alerting is centralized on top of it so that a pattern spanning two sources raises one alert to one place. The alerting thresholds are tuned on a defined cadence rather than set once, because an alert stream nobody can read is the same as no alerting at all. Time synchronization uses more than one source: at least two reference sources are configured wherever an asset supports it, so losing one does not silently leave the estate drifting. WHAT IS WATCHED includes people as well as machines: the activity of personnel and their use of the organization’s technology are monitored against what has been authorized for them and against what the organization has told them is monitored, so misuse and a compromised account surface from the same record. Analysis goes past the alert to the activity behind it - what else the same account, host or address did before and after, and whether the separate events form one sequence - so a potentially adverse event is understood rather than merely counted. And each such event is scoped before it is handed on: the estimated impact and the reach of it - which systems, which data, which accounts, over what period - is established from the correlated record and carried into the incident assessment rather than left for the responder to rebuild. | KSI-CMT-LMC, KSI-MLA-ALA, KSI-MLA-LET, KSI-MLA-OSM, KSI-MLA-RVL | A.8.15, A.8.16, A.8.17 |
| Network security controls Firewalls/segmentation and network controls restrict traffic to and from sensitive environments. The networks themselves are managed as assets with owners: what exists is documented, traffic is permitted by rule rather than by default, the rules are reviewed and the ones nobody can justify are removed, and devices connecting are authenticated rather than trusted for being on the wire. NETWORK SERVICES are treated as a separate question from the network itself, and the same question is asked whether the service is run in-house or bought: for each one - connectivity and transit, remote access, name resolution, filtering, load balancing and delivery, wireless, voice and real-time communications - the organization identifies the security mechanisms it must apply, the service levels it must meet and the management requirements that come with it, and writes them into the agreement with the provider or into the internal service definition before anything depends on it. That includes what authentication, encryption and connection controls the service enforces, the availability and capacity it commits to, who may connect and how that is decided, what it monitors and reports and to whom, and the organization’s right to verify that what was agreed is what is delivered. Services are reviewed against those terms on a cadence, and a service that cannot demonstrate them is treated as a recorded risk rather than as a working arrangement. All of it rests on a documented system and communications protection policy with supporting procedures, issued to the roles it binds, owned by a named role and reviewed on a defined cadence. THE BOUNDARY is drawn narrowly and deliberately: the number of external connections into the organization’s systems is limited rather than allowed to accumulate, each external telecommunications service terminates on a managed interface that enforces the organization’s traffic flow policy, and any exception granted to that policy is documented with the need it serves, the systems it applies to and a duration, and is reviewed and removed when the need ends. Connections BETWEEN the organization’s own system components are governed on the same terms rather than trusted for being internal: each is individually authorized, its interface characteristics, security requirements and the nature of the information communicated are documented, the conditions under which it will be terminated are stated in advance, and its continued need is reviewed on a cadence. A network connection is torn down at the end of the session it serves, or after a defined period of inactivity, rather than left open until something else closes it. Components the public can reach - the website, the mail and web gateways, the API front end, anything published to the internet - sit on subnetworks physically or logically separated from the internal network, so reaching a public component does not place the caller inside the estate behind it. WHERE INFORMATION ITSELF MAY TRAVEL is controlled as well as the traffic that carries it: the flows permitted for each classification of information - between internal systems, out to an external party, from a more trusted zone into a less trusted one, and out of the organization altogether - are decided in advance and recorded as approved authorizations, and the enforcement points are configured to those authorizations rather than to a general instruction to be careful, so an unapproved flow is blocked by a rule somebody wrote instead of being permitted because nobody wrote one. VOICE AND REAL-TIME COMMUNICATIONS are governed as a service in their own right rather than as ordinary traffic: usage restrictions and implementation guidance are written for Voice over IP and the conferencing and messaging platforms beside it, a deployment is authorized before it carries a call, and its use is monitored and controlled - because a softphone, a conferencing bridge or a SIP trunk is a path for audio out of a room and a route into the network, not only a convenience. Availability is defended as well as confidentiality: the effects of denial-of-service events - the types the organization has decided it must withstand - are limited or absorbed by controls chosen for that purpose, with the capacity and the protective mechanisms sized against a stated expectation rather than against the traffic seen so far. NAME AND ADDRESS RESOLUTION is treated as security infrastructure. Where the organization is authoritative for a namespace, its responses carry data origin authentication and integrity verification artifacts so a resolver can validate them, and the security status of each child zone is published along with the material needed to verify the chain when a child zone is operated separately. Where the organization resolves names, its resolvers request and verify those artifacts on the responses they receive from authoritative sources rather than accepting an answer because it arrived. The resolution service itself is architected for fault tolerance and separates its internal and external roles, so an outage or a compromise on one side does not carry to the other. THE NETWORK ESTATE IS MAINTAINED as an asset in its own right. Devices run releases the vendor still supports, and their versions are reviewed on a defined cadence so end of support arrives as a planned replacement; their configuration is held as version-controlled infrastructure-as-code and their management interfaces are reached only over protocols that authenticate and encrypt the session; authentication, authorization and accounting for network access are centralized rather than held device by device; and the protocols used for management and for carrying traffic are chosen from those still considered sound rather than from those the equipment happens to default to. Architecture diagrams and the supporting network documentation are maintained and reviewed on a defined cadence rather than drawn once at build. Name resolution is pointed somewhere trusted: assets resolve through resolvers the organization controls or has decided are reputable, rather than through whatever a network hands them. DETECTION AND ENFORCEMENT sit in the path as well as beside it. Intrusion detection watches network traffic for malicious activity; intrusion prevention blocks it where the organization has decided blocking is appropriate; traffic is filtered at the application layer through a proxy, application-layer firewall or gateway where a port-and-address decision is not enough; and access is controlled at the port a device connects to, the device authenticating by 802.1X or an equivalent before it is on the network rather than being trusted for having reached a socket. The documentation is kept as an explicit REPRESENTATION of what is authorized: the network communication the organization permits, and the data flows inside its own estate and across its boundary to external parties, are drawn and maintained as a current picture rather than reconstructed from firewall rules when somebody asks for one. The networks and the network services running on them are monitored on the same terms, so a potentially adverse event - an unexpected flow, a service behaving unlike its baseline, traffic to somewhere nothing should be talking to - is found by watching rather than reported by its consequences. TRUSTED AND UNTRUSTED is a boundary with two directions rather than one: traffic entering the estate from any network the organization does not control is admitted only where a rule allows it, traffic leaving for such a network is restricted on the same terms rather than permitted for having originated inside, and a packet arriving from outside that claims an internal source address is discarded at the boundary rather than routed on the strength of what it says about itself. | KSI-CNA-MAT, KSI-CNA-RNT, KSI-CNA-RVP, KSI-CNA-ULN | A.8.20, A.8.21, A.8.22 |
| Secure architecture & engineering principles The organization writes down the engineering principles its systems are designed to, and applies them to everything it builds or significantly changes rather than to whatever the architect of the day remembers. The principles state positions the design has to answer: defense in depth, so no single control failing exposes the system; least privilege for every identity, service and process; secure defaults, so an unconfigured system is safe rather than open; failing into a secure state rather than an available one; minimizing attack surface by removing rather than protecting what is not needed; validating at each layer rather than trusting a caller because it is internal; separating duties in the design so no single component or person can complete a sensitive action alone; and designing for the assumption that any component may be compromised. Each principle carries enough guidance to be applied by an engineer without asking what it means. Designs are reviewed against the principles at a defined point before build starts, a deviation is recorded and approved with its reason rather than argued informally, and the review reaches systems that are bought or outsourced as well as ones written in-house. The principles are revisited as threats, platforms and technologies change, and the changes are pushed to the teams designing against them. The output is an ARCHITECTURE, not only a set of positions: security and privacy architectures are developed and maintained for the organization’s systems, describing how the requirements are met, how the information is protected through the whole of its life, how privacy risk to individuals is minimized by design, and how the design assumptions and dependencies hold - and they are reconciled with the organization’s enterprise architecture rather than drawn beside it, reviewed and updated as the environment changes, and reflected in the acquisition and procurement decisions taken from them. Four specific separations follow from the principles and are stated so they cannot be argued away. User-facing functionality, including user interfaces and services, is kept physically or logically separate from system management functionality, so an administrative capability is not reachable from the surface every user already has. Unauthorized and unintended transfer of information through SHARED SYSTEM RESOURCES is prevented, so memory, storage, registers, caches and buffers released by one user or process do not deliver residual data to the next. Each executing process is maintained in a separate execution domain, so one process cannot read or modify another’s state. And memory is protected by mechanisms that prevent unauthorized code from executing in it. Two of the principles are stated concretely because they are the ones a design quietly skips. MEDIATION: every operation a user asks for is validated at the point it is acted on rather than trusted because an earlier screen allowed it, and input is subjected to explicit, documented checking for size, data type, and the ranges or formats the operation accepts - the position being that user input is never trusted, including input that arrived by way of the organization’s own front end. REUSE OVER REINVENTION: the security-sensitive components of an application - identity and authentication, authorization, cryptography, audit logging - are built on vetted platform mechanisms, established libraries or services rather than written afresh, and only standardized, currently accepted and widely reviewed cryptographic algorithms are used, because a bespoke implementation of a solved problem is where the implementation errors are. | KSI-CNA-DFP, KSI-CNA-MAT, KSI-CNA-ULN, KSI-SVC-PRR | A.8.27 |
| Secure configuration & baselines What a secure configuration IS is defined before anything is deployed, and drift away from it is detected rather than discovered. Baselines are written for each class of hardware, software, service and network component the organization runs, covering the settings that carry security consequence: vendor defaults changed, default and unused accounts removed or disabled, unnecessary services, ports, features and sample content removed, logging and time settings enabled, authentication and encryption settings set, and the administrative interfaces restricted. Baselines are derived from a stated hardening source and from the organization’s own requirements, are version-controlled, are approved before use, and are reviewed when the product changes, when a new threat makes a previous setting inadequate, or on a defined cadence. What is actually deployed is compared against the baseline automatically and on a schedule, and a difference raises an alert to somebody who either corrects it or records an approved exception with a reason, an owner and an expiry. Configuration is applied by an automated, repeatable mechanism wherever possible, so a rebuilt system comes back hardened rather than as somebody remembers it. A change to a baseline goes through the change process, and the record of what a system’s configuration was at a given time is retained - previous versions are kept deliberately and for a stated number of generations, so a system can be rolled back to a known-good configuration rather than rebuilt from memory. The settings themselves are set to the most restrictive mode consistent with what the system has to do, and any deviation from that is approved, documented with the operational requirement behind it, and monitored rather than left implicit. LEAST FUNCTIONALITY is the rule the baseline expresses: a system provides only the capabilities its purpose requires, and the functions, ports, protocols, software and services that purpose does not require are disabled, restricted or removed. That is not a one-time act at build - the estate is reviewed on a defined cadence to find functions, ports, protocols, software and services that have become unnecessary or are insecure by current understanding, and what the review finds is disabled or removed. Two artifacts sit alongside the baselines. A configuration management plan states the roles and responsibilities for the process, the process itself, the configuration items that are placed under configuration management and how they are identified, and it is protected against unauthorized disclosure and modification in its own right because it maps the estate. And the documentation for each product is obtained or written at acquisition rather than assumed to be findable later, in two parts. Administrator documentation covers secure configuration, installation and operation, the effective use and maintenance of the security and privacy functions, and the known vulnerabilities in how the administrative and privileged functions are configured and used. User documentation covers the security and privacy functions a user can reach and how to use them effectively, the ways of interacting with the system that keep it secure and protect individual privacy, and what the user is responsible for. Both are kept current, protected, and distributed to the administrators and users who need them. Where documentation is unavailable or does not exist, the attempts to obtain it are recorded and the organization takes a decided response - writing what it needs itself, or treating the gap as a risk. The process itself is a document and not only a set of files: how a baseline is written, approved, applied and reviewed is recorded and re-approved on a defined cadence and whenever a change to the organization would alter it - and network infrastructure carries its own such process rather than being covered by the one written for servers and end-user devices, because the devices, the settings and the people who change them are different. HOW assets are managed is part of the configuration: configuration is held as version-controlled infrastructure-as-code wherever the platform allows, and administrative interfaces are reached only over protocols that authenticate and encrypt the session - an insecure management protocol is not used unless it is operationally unavoidable, and then as a recorded exception. The hardening source is industry-recommended rather than invented here, and the templates reach the whole of an application’s infrastructure - the operating system under it, the database, the web server, the container image, and the platform and software services it is assembled from - so software written in-house cannot weaken the configuration it is deployed onto. The written process is issued as well as held: the hardening policy and the procedures under it go to the people who build, deploy and maintain the components they govern, so the standard is in use by the people it binds rather than merely current in a repository, and each part of it names the role answerable for carrying it out. | KSI-CMT-RMV, KSI-CNA-DFP, KSI-CNA-EIS, KSI-CNA-IBP, KSI-MLA-EVC, KSI-SVC-ACM | A.8.9 |
| Secure software development Security is built into how software is specified, written, reviewed, tested, released and maintained, and the lifecycle is defined so that a team knows at which point each security activity happens rather than fitting them in where there is room. Underneath that sits a secure coding standard the organization actually maintains: it exists for each language and platform in use, is based on a recognized source plus what the organization’s own defects and incidents have taught it, and is versioned and dated. It applies at three points. Before coding - developers are competent in it and trained on it, the environment is set up so the insecure option is the harder one, and the components and frameworks a project may build on are chosen and pinned rather than picked up. During coding - it gives concrete direction on the failure classes that actually recur: untrusted input and injection, output encoding, authentication and session handling, authorization decided on the server, memory and resource handling, error handling that does not disclose internals, use of cryptographic functions, and keeping credentials and keys out of source. After coding - code is read by somebody other than its author before it merges, and automated analysis runs in the pipeline rather than on request. Third-party and open-source components are held to the same standard, kept at supported versions, and reviewed when a new advisory affects them. Legacy code is brought up to the standard as it is touched rather than exempted permanently, and the standard itself is revised as languages, platforms and attacks change. Where development is performed by a supplier, the same expectations are conditions of the engagement: the developer follows a documented development process that explicitly addresses security and privacy requirements, identifies the standards and tools it uses and the specific options and configurations of those tools, and documents and manages any change to the process or the tooling during development - and the organization reviews that process, those standards, those tools and their options on a defined cadence to confirm they still satisfy what it requires. The developer also performs a criticality analysis at the decision points in the life cycle the organization has named, at the level of rigor it has specified, so the components and functions that matter most are identified while the design can still change. Developer training is given at least annually rather than once at induction, and is designed to build a security culture in the team rather than to record a completion. And a fixed vulnerability is not a finished one: ROOT CAUSE ANALYSIS is performed on security vulnerabilities found in the organization’s own code, so the underlying issue that produced the defect - a missing validation habit, an unsafe helper, a design that made the wrong thing easy - is identified and addressed rather than the team moving from one individual fix to the next. How well the practices are working is MONITORED across the lifecycle rather than assumed from their existence: the security activities at each stage carry a measure somebody actually reads - what the automated analysis found and how quickly it was acted on, what code review caught, what reached production and had to be fixed there - and the results are reviewed for what they say about the process, so the standard is adjusted for the defects it is not catching. The data management process is issued as well as written: it goes to the people who create, hold and handle each category, it names the role answerable for each category rather than leaving ownership collective, and it is kept current so that what people work to is what the organization has decided rather than what it decided the last time somebody looked. How these protections are applied is written down as procedure rather than left as intent: which paths and which stores are in scope, which protocols, algorithms, key strengths and certificate sources are acceptable and which are no longer, how an exception is requested and approved and for how long, and who owns each decision - kept current as the cryptography ages, issued to the engineers and administrators who configure it, and reviewed on a defined cadence so that what people build to is the current standard. The arrangements are documented as policy and operating procedure rather than carried in the heads of the people who set them up: what is deployed on which system types, the update and scan cadences, who administers the tooling, who acts on a detection and within what time, and how a system type judged not to be at risk is re-evaluated - kept current, in use by the people it binds, known to them, and owned by named roles. All of it is documented rather than customary: the lifecycle, the standard and the activities hung off them are written down, kept current as the platforms and the attacks move, in use by the teams they bind rather than filed, known to the engineers doing the work, and owned by named roles rather than by the team collectively. How the organization tests itself is written down as a regime rather than practiced as a set of habits: which tests run against which scope, at what frequency, who performs each of them, who receives the results, and what has to happen to a finding and by when - documented, kept current, in use, known both to the people who run the tests and to those who act on what they find, with the roles and responsibilities for each assigned rather than assumed. The decisions are carried by an access control system rather than by convention: it reaches every component in scope, it resolves each request against the permissions assigned to the individual, application or system making it - assigned from the job classification and function the organization recorded, not from what the requester asked for - and it is set to refuse by default, so a request no rule permits is denied rather than allowed because nobody wrote a rule against it. Where an obligation sets a floor under the history rather than leaving the period to the organization - a stated number of months of records, of which a shorter recent window has to be immediately available to query rather than restorable from an archive - the store is configured to that floor, the fast window is sized for it, and the restore path for the older portion is tested rather than assumed to work. THE ALERTING REACHES THE OTHER CONTROLS and not only the logging pipeline: a failure of any control the organization has designated critical - the network filtering, the intrusion detection, the anti-malware, the change-detection mechanism, the physical and the logical access systems, the segmentation that bounds a sensitive environment - is detected promptly rather than at the next review, raised to a named role, and worked under a documented response that records what failed, what caused it, how long it was down, what was done to protect the environment while it was, and what was changed so it does not happen again. | KSI-CMT-VTD, KSI-PIY-RSD | A.8.25, A.8.28 |
| Vulnerability management Regular scanning, prioritization, and remediation of vulnerabilities across systems and applications, fed by current information about threats and weaknesses collected from outside the organization as well as from its own scans - vendor and industry security advisories for the software actually in use, and the threat feeds, bulletins and sector reporting that describe how systems like these are being attacked now - which is gathered continuously rather than at the next scan, evaluated for whether it applies here, and used to decide what is looked for and what is fixed first. The set of vulnerabilities the scanner actually looks for is updated on a defined cadence and whenever new ones are identified and reported, so a scan reflects what is known today rather than what the tool shipped with. Scans that need to see inside a system are given the privileged access to do so, granted deliberately to the scanning activity for the components that require it rather than left to run blind and report clean. Whether a fix is actually present is confirmed by automated mechanisms that report, per component, which security-relevant software and firmware updates are installed - so remediation is evidenced by the estate rather than by a closed ticket. The organization also runs a PUBLIC intake: a reporting channel that anybody outside the organization can find and use to report a vulnerability they have discovered in its systems or products, with a stated scope, a stated way to report, an acknowledgment, and a route into the same triage and remediation process everything else uses. All of this rests on a documented system and information integrity policy with supporting procedures, issued to the roles it binds, owned by a named role and reviewed on a defined cadence. The scanning and the fixing are each defined rather than assumed. Internal assets are scanned automatically on a defined cadence, both with credentials and without, because the two find different things - one shows what is installed, the other shows what somebody with no account can see. Externally exposed assets are scanned on their own cadence, which is at least as frequent, because they are reachable by everyone. Patching is automated for operating systems and, on the same terms and cadence, for the applications running on them, so an application left to be updated by whoever notices is not the gap. Remediation runs to a documented, risk-based strategy - what is fixed first, within what period, and who may approve an exception - reviewed on a defined cadence rather than written once. The public intake is governed by a written vulnerability handling policy that names how to report, who is responsible for handling a report, and the steps from intake through assignment and remediation to remediation testing, with reports tracked in a system that records a severity rating and the timing of identification, analysis and remediation, so how long the organization takes is a measured number rather than an impression. | KSI-PIY-RVD, KSI-SCR-MON | A.5.7, A.8.8 |
| Resilience & Continuity | ||
| Backups Regular, tested backups of critical data and systems with defined retention, each one a RETRIEVABLE EXACT COPY of the data it protects - complete and restorable, not a partial or lossy snapshot - including a copy taken before equipment holding that data is moved. Backup information is tested on a defined cadence to verify that the media are still reliable and the information still has its integrity - a restore actually performed, not a job that reported success - and it is protected by cryptographic mechanisms so a copy obtained by somebody who should not have it discloses nothing and cannot be altered undetected. Copies are held somewhere other than where the original lives: an alternate storage site is established, with the agreements needed to store backups there and to retrieve them when they are wanted, carrying security controls equivalent to those at the primary site rather than weaker ones because it is only a copy. The alternate site is chosen far enough from the primary that the same fire, flood, outage or regional event is unlikely to take both, and the organization identifies in advance the problems that would make the site hard to reach during a wide-area disruption - roads, transport, staff availability, network dependency - and states explicit mitigation actions for each rather than discovering them on the day. The recovery itself is a documented process and not only a schedule: it states which assets are in scope for recovery, the order in which they are brought back, and how the backup data is protected while it waits, and it is reviewed and updated on a defined cadence and whenever a change to the organization would alter it. Recovery data carries protection EQUIVALENT to the data it copies rather than weaker protection because it is a copy. And at least one instance of it is ISOLATED - held offline, off-site, or in a separately controlled service, out of reach of the credentials and the network paths that operate the live environment - so an event that reaches production does not also reach the copy that would undo it. Verification is performed at the point of USE as well as on the cadence: before a backup or any other restoration asset is relied on to bring a system back, its integrity is checked against the value recorded when it was taken and the check is logged - so a restoration does not carry corrupted or tampered data into a system that has just been cleaned. | KSI-RPL-ABO, KSI-RPL-TRC | A.8.13 |
| Business continuity & disaster recovery BC/DR plans with defined RTO/RPO, tested periodically AND REVISED on what the testing finds and on what has changed since, to restore service after disruption - including how the critical processes that protect sensitive data keep running while the organization is operating in emergency mode, and an assessment of how critical each application and data set is, which is what sets those recovery targets and the order in which things come back. The disruptions the organization could actually face are identified and a response chosen for each, rather than one plan written against one scenario, and the loss that would still remain after those responses is quantified and put to a deliberate decision - accepted, reduced further, or transferred, including by insurance - so exposure to a disruption is something somebody chose rather than something nobody priced. The level of information security to be MAINTAINED while the organization is disrupted is decided in advance rather than allowed to fall to whatever the emergency leaves standing: for each control that cannot run in the degraded state, a compensating measure is defined for the period, and where none is available the exposure is accepted deliberately and for a stated maximum duration. The alternate site, the standby service and the emergency working arrangements carry protection equivalent to normal operations - the same access rules, the same logging, the same handling of sensitive information - and the security of those arrangements is exercised in the same tests rather than assumed to have been inherited. Restoring normal operation includes restoring the controls that were relaxed, and confirming that they are back on. The program rests on a documented contingency planning policy with supporting procedures, issued to the roles it binds, owned by a named role and reviewed on a defined cadence. The plan is not written alone: it is developed in coordination with the organizational elements responsible for the related plans - incident response, crisis management, physical security, occupant emergency, supply chain - so the plans agree about who does what rather than each assuming the others, and the testing is coordinated with those same elements for the same reason. What has to come back is named rather than implied: the essential mission and business functions are identified, the critical system assets and components that support them are identified through a deliberate criticality analysis performed at defined points in the system’s life rather than once at the start, and the plan states the time within which each essential function will be resumed after the plan is activated. Testing runs on a defined cadence using methods chosen to establish the plan’s effectiveness and the organization’s readiness to execute it, and the results are reviewed and corrective action taken. People are trained for the role the plan gives them: within a defined period of being assigned it, again when the system or the plan changes materially, and on a defined cadence thereafter, with the training content revised for what the exercises and the incidents showed. Where a system is transaction-based, recovery includes the transactions themselves - the mechanisms that let in-flight work be rolled back or replayed to a consistent point, so recovery does not mean a service that is up over data that is half-written. The criticality analysis names the objectives, capabilities and services that parties OUTSIDE the organization depend on or expect from it - customers, regulators, and the organizations it is itself a supplier to - and what the organization commits to restoring, and how quickly, is communicated to them rather than held internally. Recovery is entered deliberately rather than drifted into: the criteria for initiating it are set in advance and applied to the incident in front of the responders, and the recovery actions are then selected, scoped, prioritized and performed against the plan instead of improvised in the order things are noticed. What normal looks like afterwards is a decision too - the essential mission functions and the risk picture the incident has just changed are both considered when the post-incident operating norms are set, so the organization does not return to a posture the incident has already disproved. The end of recovery is declared against stated criteria by the role authorized to declare it, and the recovery documentation is completed at that point rather than left open behind a service that is back up. | KSI-RPL-ARP, KSI-RPL-RRO, KSI-RPL-TRC | A.5.29, A.5.30 |
| Incident response A documented, tested plan to detect, triage, contain, remediate, and communicate security incidents, and to mitigate - so far as is practicable - the harmful effect of a use or disclosure of personal data the organization knows breached its own policies or the law. Each incident is recorded together with its outcome - what happened, what was done about it and how it ended - as a record of that incident, which is a different artifact from the plan being documented. The mitigation duty runs to violations by the organization itself AND to violations by the processors, vendors and other parties handling that data on its behalf: the plan reaches an incident somebody else caused with the organization’s data, so learning of one triggers the same containment and remediation as an incident inside its own walls rather than a request that the other party deal with it. Where an incident carries a duty to tell someone outside the organization, the plan discharges it on the clock the applicable law sets - and, where the organization has itself committed to a timeframe for telling people, on that commitment too, whether or not a statute stands behind it - rather than whenever the investigation happens to conclude: whether an incident is notifiable is decided against written criteria rather than argued after the fact, the regulator or supervisory authority is notified inside the deadline that regime states and inside any shorter or additional timeframe the organization has committed to, the people whose data is affected are told where the risk to them warrants it and, independently of that threshold, wherever the organization’s own privacy commitments say they will be told - so individual notification is never conditioned solely on a statutory risk test - and any other party that law or those commitments require to be notified is told on the same terms, and where a deadline is missed the notification itself explains the delay instead of passing over it. What a notification carries is fixed in advance rather than composed under pressure: to a regulator it describes at least the nature of what happened, including where possible the categories and the approximate number of people affected and of records involved; names a contact point - the data protection officer where there is one, otherwise whoever can answer - from whom more can be obtained; describes the likely consequences; and describes the measures taken or proposed to address it, including where appropriate the measures that will mitigate its adverse effects. Where all of that cannot honestly be given at once, it is given in phases without further undue delay rather than held back until the picture is complete, and each phase says what is still outstanding. The communication to the people affected describes what happened in clear and plain language and carries the same contact point, likely consequences and measures. Every compromise of personal data is documented whether or not it turned out to be notifiable - the facts of it, its effects, and the remedial action taken - in enough detail that a regulator reviewing the file can verify for itself that the notification decision was the right one. Recovery is part of the plan rather than something that follows it: service and data are restored to a state the organization has established is clean, the restoration is verified before the system is handed back to use, the cause is determined rather than inferred from the symptom, and the weakness the incident exposed is fixed - with the plan itself updated for what the incident showed about it. Between the report and the response sits an assessment step that is a duty of its own: every reported event is assessed against written categorization and prioritization criteria by people competent to apply them, and the decision - whether this event is an incident, and at what severity - is recorded with the reasoning, so two assessors reach the same answer and an event judged not to be an incident is a decision somebody made rather than a report that went quiet. Learning is treated as a duty separate from fixing the incident in front of you: the types, volumes and costs of incidents are quantified and reviewed as a set for what the pattern says, and what is learned is pushed back into the controls, the risk assessment, the awareness material and the assessment criteria themselves rather than staying in the report of the incident that produced it. The plan is a documented incident response policy with supporting procedures, issued to the roles it binds, owned by a named role, and reviewed and updated on a defined cadence. The people the plan assigns roles to are trained for them: within a defined period of taking the role, again when the system or the plan changes in a way that affects it, and on a defined cadence thereafter, with the content revised for what exercises and real incidents have shown. The capability is TESTED rather than assumed - on a defined cadence, using tests the organization has chosen for the purpose, such as a tabletop, a walkthrough, a simulation or a live exercise - and that testing is coordinated with the organizational elements that own the related plans, incident response and contingency planning in particular, so the two do not each assume the other. Handling and reporting are supported by automated mechanisms rather than run by hand at the worst moment: detection, triage, tracking, evidence collection and the routing of a report are automated so far as the organization’s systems allow, and the reports that must go outside are produced and sent by mechanism rather than composed under pressure. The roles the plan assigns are named across the functions an incident actually needs and not security alone - legal, IT, information security, facilities, communications, human resources, the responders and the analysts - and the assignment is reviewed on a defined cadence. So are the CHANNELS: a primary and a secondary mechanism for communicating and reporting during an incident are chosen in advance, on the understanding that the ordinary one may be the thing that is unavailable or compromised, and both are reviewed on the same cadence. The plan reaches the parties outside the organization that an incident actually involves: the suppliers and other third parties whose services, staff or systems would be part of the response are named in it, take part in the planning and the exercises, and are called on during response and recovery on terms agreed in advance rather than negotiated during the event. ESCALATION is a defined step and not a judgment call - the plan states the conditions under which an incident is escalated or elevated, whether by severity, by elapsed time, by the functions it has reached or by the obligations it triggers, who it goes to at each step, and what changes when it gets there. The analysis establishes what actually took place during the incident as well as why it happened, and the incident’s magnitude - how many systems, records and people it reached, and over what period - is estimated as the investigation proceeds and then VALIDATED against the evidence rather than left at the first number anybody said out loud. Notification runs to internal stakeholders as well as external ones, so the functions inside the organization that have to act on an incident are told on the same defined terms as the parties outside it. And containment is followed by ERADICATION as a separate act: the malicious code, the unauthorized access and the persistence left behind are removed and their removal is confirmed, so a contained incident is not mistaken for a finished one. | KSI-INR-AAR, KSI-INR-RIR, KSI-INR-RPI | A.5.24, A.5.25, A.5.26, A.5.27 |
| Redundancy of processing facilities Redundancy is built into the systems and facilities that process and hold information, to the level the availability requirements for each service actually demand - and those requirements are stated per service first, so redundancy is engineered against a target rather than bought to a feeling. Single points of failure are identified deliberately, component by component and site by site, and each is either removed, duplicated, or accepted as a risk somebody has signed for. Where duplication exists, the standby is sized to carry real production load rather than to exist, and is kept at the same patch, configuration and data currency as the primary so failing over does not mean failing back to an older state. Failover is TESTED on a schedule and after significant change, in a way that actually moves the service rather than confirming that a switch exists, and the time it took is measured against the target. The redundant path is monitored in its own right, so a failed standby is found before the primary needs it. Where redundancy spans providers or regions, the dependency they share - a common network, a common identity service, a common control plane - is identified, because two copies behind one dependency is one copy. An ALTERNATE PROCESSING SITE is established for the essential functions that need one, with agreements permitting the transfer and resumption of those functions within the time the organization has set, equipped and configured to run them - or with the arrangements in place to equip it in time - and carrying security and privacy controls equivalent to those at the primary site. It is separated from the primary site far enough that the same regional event is unlikely to reach both; the problems that would make it hard to reach during a wide-area disruption are identified in advance with explicit mitigations; and the agreements behind it carry priority-of-service provisions consistent with the availability targets, so the organization is not one customer among many at the moment it needs the site. ALTERNATE TELECOMMUNICATIONS are arranged on the same terms: services that permit the essential functions to resume within the defined time when the primary carrier is unavailable, agreements that include priority-of-service provisions - including registering for the national priority scheme where the organization qualifies and needs it - and a check that the alternate does not share a single point of failure with the primary, which is the usual way two carriers turn out to be one. | KSI-CNA-OFA | A.8.14 |
| Third-party Risk | ||
| ICT supply chain security Security expectations follow the technology the organization buys, not just the supplier it bought it from. Before acquisition, the security requirements for an ICT product or service are stated as conditions of purchase, and the supplier is required to propagate them to the sub-suppliers, components and services it in turn depends on rather than absorbing them at its own boundary; where the supplier will not or cannot, that is a recorded acceptance decision rather than a silent gap. The organization establishes what a delivered component is actually made of - the third-party and open-source software inside it, and where it came from - and keeps that record current enough to answer whether a newly published vulnerability affects it. Delivery is verified rather than assumed: the component received matches the one ordered, comes from the expected source, and its integrity can be checked. Through life, the supplier’s security advisories and the product’s support status are watched, so end-of-support arrives as a planned migration rather than as a discovery; and the sourcing risk itself - a supplier failing, being acquired, or ceasing to supply a component with no substitute - is assessed for the components the organization cannot operate without. That assessment is performed as a SUPPLY CHAIN RISK ASSESSMENT in its own right, covering the systems, components and services the organization depends on, and is updated on a defined cadence and whenever the supply chain materially changes - a new supplier, a change of ownership, a relocated manufacturing or support function, a substituted component. The program is written down: a documented supply chain risk management policy with supporting procedures under a named owner, reviewed on a defined cadence; and a supply chain risk management plan covering the whole life of the system from acquisition through operation to disposal, protected from unauthorized disclosure and modification because it maps the organization’s dependencies, and reviewed and updated on a defined cadence. A team is stood up to lead and support that work, drawn from the functions that actually decide - procurement, engineering, security, legal - rather than left as one person’s additional duty. Procurement is one of the controls: the acquisition strategies, contract tools and methods the organization uses are chosen to guard against, identify and mitigate supply chain risk, and agreements with suppliers and other entities in the chain set out what each party must notify the other of, disclose, or otherwise do when something changes or goes wrong. Delivered systems and components are inspected at defined times or on defined events to detect tampering, and an anti-counterfeit policy with supporting procedures exists to detect and prevent counterfeit components entering the estate, to report any that are found to the source, the authorities and the affected parties, and to train the personnel who receive, install and maintain equipment in what a counterfeit looks like across hardware, software and firmware. When an incident happens, information about it is shared with the providers of the affected products and services and with the other parties across the supply chain who need it, rather than treated as internal. For software the organization develops, that record is maintained as a BILL OF MATERIALS covering the third-party components in use and the ones a project intends to adopt, with the risk each component carries noted against it, and it is evaluated on a defined cadence - at least monthly - to pick up a change in a component or the loss of its support rather than waiting for an advisory to arrive. The program is not run beside the others: what the supply chain risk assessment finds enters the organization’s cybersecurity and enterprise risk registers, is assessed on the same criteria as every other risk, and feeds the same improvement process - so a supply chain finding competes for attention on equal terms instead of sitting in a document only procurement reads. How well the practices are working is monitored across the whole life of each technology product and service - at selection, at delivery, in operation and at retirement - rather than assessed once at purchase. And the suppliers the organization could not operate without are assessed BEFORE acquisition, against stated criteria and with the result recorded, so criticality is established while there is still the option not to buy. | KSI-SCR-MIT, KSI-SCR-MON | A.5.21 |
| Third-party / vendor risk management Due diligence, contractual safeguards, and ongoing monitoring of vendors that handle your data: the agreement obliges the vendor to comply in its own right with the security requirements that apply to it - an absolute standard, not a promise to match whatever you happen to do - to pass those obligations down to any subcontractor it brings in BY ENTERING INTO a contract or equivalent written arrangement with that subcontractor rather than by merely requiring equivalent practice of it, and to report to you, within a stated time, security incidents it becomes aware of and confirmed breaches of your data. Where a contract is not the instrument available, an equivalent written arrangement carrying the same obligations discharges the duty. The same obligations, together with the separation that keeps a related organization out of data it is not entitled to, are written into the governing document of any other arrangement that puts your data in the hands of a sponsor, parent, affiliate or plan. Diligence is not confined to security where the relationship warrants more: for suppliers significant enough to matter, the organization states the standards of conduct it expects of them - how they behave commercially and how they treat the environment around their operations - and screens candidates and incumbents against those stated expectations as part of the same selection and monitoring cycle, rather than accepting a signature on a code as evidence of it. Where the vendor handles personal data, the agreement binds it to privacy obligations no weaker than the commitments the organization has itself made about that data - the purposes it may be used for, the limits on passing it on further, and the help the organization needs in order to answer the requests individuals make about it - and the reporting duty above reaches a suspected as well as a confirmed compromise of that personal data, on the same stated clock. Which requirements apply to a given supplier is decided by the TYPE of relationship rather than by one clause set issued to everyone - what data it touches, what access it holds, whether it can affect the organization’s own service, and what it would cost if it failed - and the requirements are agreed and recorded before access begins rather than negotiated after go-live. Once the relationship is running, what the supplier actually delivers is reviewed against what was agreed on a stated cadence: the service records, the security reports and assurance the agreement entitles the organization to, the incidents it has declared, and the findings of any audit or test right the organization holds - exercised rather than merely retained. A change on the supplier’s side is managed as a change rather than discovered - a new subcontractor, a new location or jurisdiction, a change of ownership, a material change to the technology or to the people delivering the service is notified in advance under the agreement, assessed for what it does to the risk, and approved or refused before it takes effect. ACQUISITION is governed as its own act, under a documented system and services acquisition policy with supporting procedures, owned by a named role and reviewed on a defined cadence. When a system, a component or a service is bought, the contract states the security and privacy requirements it must meet - the functional requirements, meaning what the controls have to do; the strength requirements; the assurance requirements, meaning what evidence the supplier must produce that they work; the documentation the supplier must deliver and how it must be protected and distributed; the description of the development environment and of the environment the product will run in; and the acceptance criteria the delivery is measured against - all stated in the solicitation before a supplier is chosen rather than negotiated after award, and all expressed in terms of the applicable laws and standards. The supplier is required to describe the functional properties of the controls it will implement, and to provide design and implementation information for those controls at a level of detail the organization has specified, so the organization can judge them rather than take their existence on trust. It is also required to identify the functions, ports, protocols and other services the delivered product intends to use in the organization’s environment - and, for an external service provider, the ones its service requires - so an integration does not open a path nobody asked for. The program has three artifacts of its own. An INVENTORY of service providers lists every one the organization knows of, records the classification given to it and names the person inside the organization who owns the relationship, and is reviewed on a defined cadence and whenever a change to the organization would alter it. A POLICY governs the whole cycle - how providers are classified, how the inventory is kept, how they are assessed, how they are monitored and how they are decommissioned - owned by a named role and reviewed on the same terms. And a CLASSIFICATION is applied to each provider against stated criteria such as the sensitivity and volume of the data it holds, the availability the organization depends on it for, the regulation that reaches it, and the risk that remains after the controls in place - reviewed rather than assigned once. DECOMMISSIONING is performed rather than allowed to lapse: when a relationship ends, the user and service accounts are deactivated, the data flows into and out of the provider are terminated, and the organization’s data held in the provider’s systems is disposed of and the disposal evidenced. Who does what is settled before the relationship starts and written down on both sides: the cybersecurity roles and responsibilities of the organization, of the supplier, and of the customers and partners the arrangement reaches are established, communicated to each of them and coordinated between them, so a duty is not left in the gap where each party assumed the other held it. Planning and due diligence come before the agreement rather than after it - what the relationship would expose, what the candidate’s security actually looks like, and what would have to be true before it starts are established while declining is still an option. The risk a supplier carries is then held as a record rather than as an impression: understood, written down, prioritized against the other suppliers, assessed on a stated cadence, responded to with an owner and a date, and monitored for the whole life of the relationship instead of at onboarding only. The provider inventory records the SERVICES each one actually provides as well as its name, so what the organization has placed outside itself is answerable from the list. Where a PROCESS itself is provided from outside, it stays inside the management system’s control rather than leaving it: the controls the organization intends to apply to the external provider and the controls it intends to apply to the resulting output are defined separately and both are applied, because a well-governed supplier can still ship a nonconforming output. What the arrangement could do to the organization’s own ability to consistently meet its customers’ requirements is considered when those controls are set, and the verification or other activity necessary to establish that what arrives meets requirements is determined in advance and carried out rather than inferred from the supplier’s own assurances. Where a regime requires the CHAIN OF CUSTODY of a device to be established before it enters the environment, the organization documents and maintains that custody, replacement devices included, so the integrity of what arrives is demonstrated rather than assumed. | KSI-SCR-MIT | A.5.19, A.5.20, A.5.22 |
| People & Culture | ||
| Security awareness training Ongoing security and data-handling awareness training for all personnel, with completion tracking, and periodic security updates - reminders, bulletins and alerts - issued to the workforce between training cycles. New joiners are trained within a defined period of starting, anyone whose work is affected is retrained within a defined period after a material change to the policies or procedures, and every completion is recorded. The program itself rests on a documented awareness and training policy with supporting procedures, issued to the people and roles it binds, owned by a named role, and reviewed and updated on a defined cadence rather than at whatever point somebody notices it is stale. The curriculum names two threats explicitly, because both are answered by a person rather than by a system. The first is INSIDER THREAT: what the potential indicators look like - unexplained access outside a role, bulk copying, hostility after a disciplinary or a passed-over promotion, working around a control rather than raising it - and where to report a concern about a colleague, without the reporter being asked to conclude anything. The second is SOCIAL ENGINEERING AND SOCIAL MINING: the phishing message, the pretext phone call, the urgent request from an apparent executive, the person following somebody through a door, and the slower pattern of harmless-seeming questions that assembles into an answer nobody would have given at once - together with the instruction to report both the attempts that worked and those that did not. The curriculum is stated as a set of topics rather than left to whoever assembles the material. AUTHENTICATION: how multi-factor authentication works and why it is required, what makes a passphrase strong, and how credentials are stored and never shared. DATA HANDLING: how to identify sensitive information and how to store, transfer, archive and destroy it, together with the clear screen and clear desk habits that go with it - locking a screen on standing up, clearing a whiteboard at the end of a meeting, and putting paper and portable media away rather than leaving them out. UNINTENTIONAL EXPOSURE: the ways data leaves by accident, such as a message sent to the wrong recipient, a portable device left behind, or a file published to a wider audience than intended. INCIDENTS: how to recognize that something may be an incident and how to report it without first establishing that it is. MISSING UPDATES: how to tell that an asset is not receiving its security updates, and to report a failure of an automated patching tool rather than assume somebody is watching it. INSECURE NETWORKS: the risk of connecting to and sending organizational data over networks the organization does not control, including what is expected of a home network where people work from one. And beyond the common curriculum, ROLE-SPECIFIC training is given where a role carries specific risk - system administration, secure development, and the roles most likely to be targeted directly. | KSI-CED-RAT | A.6.3 |
Beyond the pair
Where else this work counts
A framework is lit when a shared control above also maps to it. Unlit means none of them do, which is an absence rather than a judgment about that standard.
Also reached by these 25 controls
- AI Governance Essentials not reached
- Amazon Appstore Child-Directed Apps not reached
- Apple App Store Kids Category not reached
- CIS Critical Security Controls also reached
- COPPA also reached
- ESG Essentials also reached
- EU AI Act not reached
- FedRAMP Consolidated Rules also reached
- FedRAMP Rev5 Class B also reached
- FedRAMP Rev5 Class C also reached
- FedRAMP Rev5 Class D also reached
- GDPR also reached
- Google Play Families not reached
- HIPAA also reached
- ISO 9001 also reached
- ISO/IEC 42001 also reached
- NIST AI Risk Management Framework also reached
- NIST Cybersecurity Framework also reached
- NIST SP 800-171 also reached
- NIST SP 800-53 also reached
- PCI DSS also reached
- PIPEDA also reached
- SOC 2 also reached
- SOX (Sarbanes-Oxley) Section 404 also reached
- US Employment Law - Federal Baseline also reached
Nearby pairs
- ISO/IEC 27001 and FedRAMP Rev5 Class D 55 shared controls
- ISO/IEC 27001 and FedRAMP Rev5 Class C 53 shared controls
- ISO/IEC 27001 and NIST SP 800-53 52 shared controls
- ISO/IEC 27001 and NIST Cybersecurity Framework 45 shared controls
- ISO/IEC 27001 and FedRAMP Rev5 Class B 41 shared controls
- ISO/IEC 27001 and NIST SP 800-171 35 shared controls
The thesis
Why this is one project, not two
On a crosswalk-native model, ISO/IEC 27001 mostly lights up controls you already built for FedRAMP 20x. You’re not re-uploading the same screenshot for a second audit. You apply the framework and see the genuine delta worth working. That’s the whole idea behind collect once, comply everywhere.
Next step
Add ISO/IEC 27001 to the work you already did
Apply both frameworks in one workspace and see the overlap measured against the controls you already hold.