When analyzing the resilience of a critical process, the most obvious dependencies are usually easy to identify. Electricity, telecommunications, IT applications, facilities, personnel, and suppliers almost always appear in a business impact analysis (BIA) or a resilience assessment. The problem arises when the analysis is limited to this first level. A seemingly redundant dependency may in fact rely on the same physical infrastructure. Two operators may use the same stretch of fiber-optic cable; two data centers may depend on the same electrical substation; an emergency generator may depend on a fuel supply chain that, in turn, requires power and transportation. In these cases, although redundancy exists from a formal standpoint, it does not eliminate the true point of weakness.
CISA (the Cybersecurity and Infrastructure Security Agency, a U.S. federal agency within the Department of Homeland Security (DHS)) defines critical infrastructure as highly interconnected systems and emphasizes that dependencies can be one-way or reciprocal. Energy and telecommunications are classic examples: communication systems require electricity to function, while power grids use telecommunications systems to monitor and control the transmission of energy. If one dependency is compromised, the effect can therefore spread from one system to another.
Not all addictions are physical in nature
One of the most useful contributions of CISA’s “Infrastructure Dependency Primer” is the distinction between four types of dependencies: physical, cybernetic, geographic, and logical. (The classification into physical, cybernetic, geographic, and logical dependencies dates back to the work of Rinaldi, Peerenboom, and Kelly (2001) and is taken from CISA’s Dependency Primer.) Physical dependency is usually the most intuitive. A facility requires energy, water, fuel, or raw materials to operate. Cyber dependency, on the other hand, concerns the data, networks, and information systems necessary for controlling or managing operations. A gas station, for example, may physically have the product on hand but might be unable to dispense it if its IT, accounting, or communication systems are not functioning. Geographic dependence is more insidious. Two seemingly independent systems may share the same physical “corridor” or geographic area and are therefore potentially exposed to the same fire, flood, landslide, or accident. Two fiber-optic cables from different operators may run through the same utility tunnel; a power line and a telecommunications network may cross the same bridge; and several data centers may be located within the same watershed. Finally, there are logical dependencies determined by rules, decisions, economic conditions, or the availability of personnel. A service may be technically operational but unusable due to a lack of authorizations, specific technical expertise, or the ability to deploy personnel and equipment. For a comprehensive assessment of resilience, it is therefore not enough to simply ask which infrastructure a process depends on. It is necessary to understand how it depends on that infrastructure.
From first-level dependencies to second-level dependencies
A common mistake in business impact analysis (BIA) is settling for the first answer. If a process indicates that it depends on a cloud application, that application is identified as a critical resource, and the analysis may end there. However, that application in turn depends on the cloud service provider, telecommunications, the identity provider, data center power, DNS, and a range of external services. The same logic applies to physical infrastructure. If a process depends on electricity and has a generator, the dependency seems mitigated. In reality, however, one must ask how much fuel is available, how it is replenished, where it comes from, what infrastructure is required for its supply, and to what extent the supplier can guarantee service in the event of a regional emergency. The methodological step, therefore, consists of transforming a simple list of dependencies into a chain of dependencies. For each critical resource, one must ask what it depends on to continue functioning. The same question must be asked at least one or two subsequent levels down the chain, until common infrastructure, points of concentration, or elements for which no realistic alternatives exist become apparent.
The Difference Between Redundancy and Independence
Having two resources does not necessarily mean there is redundancy. Redundancy only makes sense if both resources are not taken offline by the same event. It is therefore necessary to distinguish between redundancy and independence. Two Internet connections may be redundant from a contractual standpoint, but they are not independent if they share the same fiber-optic line. Two UPS systems may be redundant, but they may depend on the same electrical panel. Two data centers may be located in different cities but use the same identity provider or the same cloud platform. This is where the concept of common-cause failure comes into play: a single cause simultaneously compromises multiple components that were previously considered alternatives to one another. From a technical standpoint, the right question is therefore not “Do we have an alternative solution?” but “Could the alternative solution be affected by the same event that makes the primary solution inaccessible?”
A simple table can be more useful than a complex diagram
First of all, sophisticated software is by no means necessary. A simple table may be sufficient. The rows can list critical processes or services, while the columns list the infrastructure on which they depend: power, telecommunications, water, cloud, data centers, transportation, logistics providers, payment systems, and other relevant resources. The first step is to identify direct dependencies. The second step is to specify the maximum tolerable downtime (MTPD) and the recovery time objective (RTO). The third step is to list the available alternatives and verify whether they are truly independent. At this point, two very useful elements emerge: the infrastructure on which numerous critical processes simultaneously depend, and the alternatives that, upon closer analysis, share the same weakness. It is precisely here that a classic dependency matrix becomes a tool for strengthening resilience.
Search for Nodes with High Concentration
The next objective is to identify the nodes toward which numerous dependencies converge. An identity provider, an electrical substation, a single telecommunications operator, a data center, a logistics terminal, or a single fuel supplier can support countless processes simultaneously. These nodes deserve higher priority than resources that support only a single service, even if their probability of failure does not appear to be particularly high. This is in line with the approach proposed by CISA, which recommends identifying dependencies between systems and using that information to define the most effective resilience measures. Such measures may include truly independent secondary suppliers, connections to alternative systems, emergency power supplies, and operational procedures designed to maintain essential functions in the event of an outage.
Time is a technical variable
A dependency is not necessarily critical in and of itself. It becomes critical only over time. A ten-minute power outage can be fully compensated for by UPS systems and generators. A 48-hour outage can cause supply issues. A week-long outage can compromise staffing, logistics, and maintenance. Therefore, every dependency must be considered not only in relation to the availability of an alternative solution, but also to its actual autonomy. The analysis must determine how long the primary solution can remain inactive before the process is affected and how long the backup system can effectively maintain operations. The difference between these two values often makes the resilience gap much more evident than a simple indication such as “reserve available: yes/no.”
From the Diagram to the Test
A dependency analysis is useful only if it is verified. If two operators are declared independent, evidence regarding the physical routes or at least a plausible confirmation of the separation must be available. If a generator guarantees 48 hours of autonomy, it must be verified whether this autonomy was calculated based on the actual load. If an alternative supplier exists, it is necessary to know how long it would take for that supplier to actually intervene in the event of a regional incident. The most useful test often involves starting with a single piece of infrastructure and examining which services would gradually be disrupted over time. This approach allows us to observe chain reactions and identify the points at which planned solutions reach their limits. This is by no means a purely theoretical issue. NIST has reiterated that critical infrastructures —including energy, water, fuel, communications, and transportation—are interconnected and interdependent, precisely because each relies on the services of the others, both for operation and for recovery following an incident.
From a List of Resources to Mapping Resilience
A traditional business impact analysis (BIA) identifies the resources needed to execute a critical process. A more advanced dependency analysis must go further and show how these resources, in turn, depend on other infrastructure, where bottlenecks occur, and where alternatives share the same vulnerability. The value of the analysis therefore lies not in the number of dependencies identified, but in the ability to recognize those that can simultaneously put multiple services at risk and render contingency strategies ineffective. If dependency mapping succeeds in highlighting not only what we depend on, but also what our alternatives depend on, it becomes one of the most useful tools for making the transition from documentation-based business continuity to true resilience.
