The Q&A about EBIOS

Destined to be regularly enriched, those Questions & Answers (Q&A) deal with questions related to the implementation of the EBIOS method.

How to avoid the combinatorial explosion of a study?

Answer from a Club EBIOS member: “We can act on several elements so that the result corresponds to the expectations”

Studies are sometimes criticized because of the combinatorial explosion of the elements to be studied. Therefore, before or at the beginning of any study, it is necessary to wonder what the sponsor is able to accept in terms of readability.Some wish to have as much detail as necessary to treat the risks (and/or justify the measures) in a fine way.

In this case, it is possible to handle the entire combination of events and threat scenarios.

If this is not the case, here are some tips that will help you reduce the entropy of the analysis:

  • act on the presentation: keep the detailed study as a “working document” and group the risks into families in a “summary document” to facilitate decision-making. This “summary document” may only highlight the most important risks (in terms of severity and/or likelyhood), as well as those of specific interest to the sponsor;
  • act on the number of assets: to group the essential assets and/or the supporting assets in the context study. It is possible for example to adapt the level of detail of the modeling without necessarily seeking homogeneity in the model. For example, the description of supporting assets may contain both systems (for assets on which the threat analysis does not need to be detailed) and network, hardware, and software elements (for assets on which the threat analysis must be more refined);
  • act on the hypotheses: to limit the complexity of the study by reducing the combinatorics of the analysis to the only questions or justifications that one wishes to expose. To do this, it is possible to set hypotheses in the context of the study. Thus, it can be considered as a postulate that a supporting asset (or an essential asset) is protected against a type of threat (for example, “a homologation proves that the servers and workstations are sufficiently protected from all the malicious scenarios coming from ‘external attacks’). One can also consider that a risk is sufficiently covered by a certification without requiring a decomposition of this risk (for example, “the private key stored in the certified electronic chips is sufficiently protected against all threats leading to a disclosure”). It is also possible to make assumptions about residual risks for which the study is not expected to provide justification or that the study acts on the related supporting asset (for example, “the GPS is considered as no reliable, it is likely to provide bad location data “);
  • act on the decomposition into several studies: another method can consist of breaking down the studied system to transform a complex analysis into several less difficult studies to carry out. In this case, particular attention should be paid to the interfaces between these subsystems.

Is the traceability criterion useful?

Answer from a Club EBIOS member: “Two options”

Several proposals are possible to demonstrate that the problems associated with traceability are treated in an EBIOS study without considering traceability as a criterion:

  • an elegant but somewhat theoretical solution: we consider the information “traces” (or “proof”, or “log”) as an essential asset, and the traceability becomes the integrity and the availability of this essential asset. This means limiting the study to the traces that induce a feared event and to put the others out of the perimeter of the essential assets, in order to avoid the columns filled with “0”;
  • an applied solution: traceability is not a criterion, but a security measure. Being able to trace an action is a measure of both deterrence and recovery, and to consider traceability as such allows us to limit ourselves to the study of (really) feared events: we admit that not being able to trace is not really the feared event, but reduces the associated risk.

Another answer’s response: “Traceability is not a security criterion”

The security criteria are used to assess the impacts in case of reaching each of them, and in particular to study the security needs. In information security, only availability, integrity and confidentiality are considered as security criteria (see in particular ISO/IEC 2700x).They should not be confused with the topics of security measures or regulatory references. Indeed, the (false) need for traceability comes from the fact that we want to know what happened after an incident (detection measure) and/or various obligations (legal, regulatory, sectoral or security policy-related). It is therefore useless and even counterproductive to study the need for traceability.

In addition, a scale of needs and a scale of impacts related to traceability should be available. It is often by trying to build them that one realizes that it is a “desire of someone” that falls under security measures or coverage of a legal “risk”.

Finally, this would involve studying all the threats that lead to the loss of traceability! Actually, this is related to the good implementation of a security measure, which is not necessary to treat as a risk (or otherwise it should be done for encryption, access control, etc.).

However, as the study of the needs is a communication tool with business, it is possible to integrate traceability into the security criteria so that business becomes more involved in the process by seeing its point of view taken into account…

How to use scales of heterogeneous impacts?

Answer from a Club EBIOS member: “Pay attention to scales using several types of impacts”

EBIOS guide reminder: “This action [scale development] consists of creating a scale describing all possible levels of impacts, just like the scales of needs, a scale of impact levels is usually ordinal (the objects are classified in order of magnitude, the numbers indicate ranks and not quantities) and composed of several levels to classify all risks“.Therefore, it is usual to see users of the method build several ordinal scales depending on the nature of the impact (financial, legal, operations, privacy…) to estimate the severity of the feared events. The construction is then done by individually scaling each type of impact, without worrying about the consistency between the levels.

However, only a global result is used in risk maps to assess the severity of each compared to the others. Information on the nature of the impacts is lost.

To avoid misleading conclusions about the importance of risks, care must be taken to check the transverse coherence of the gradation of impacts in the scales. For example, checking that the estimatation of a level 3 impact on operations will be of the same value for the organization as the financial and legal impacts of the same level.
Where possible, the pivotal criterion (for consistency) may be the financial scale. If this is not the case (often the case), the side-by-side impacts should be presented and their importance assessed by those seeking consensus. In this case, the scales can have empty boxes (level having no equivalence for all types of impacts considered). This can be the case when one estimates the loss of human lives for example.

It is sometimes difficult for managers to establish these scales in a generic way. A good solution is then to ask the stakeholders to prioritize the feared events after identifying the impacts, and build the scales based on this estimate.

Answer from another member: “It is useful to have heterogeneous impact scales, it is an important means of communication with the business”

In addition, the ideal is in my opinion:

  • to have a scale for each type of impacts (financial, image, legal, operation, privacy…) by covering the entire spectrum of possibilities (from worst to best);
  • to present the impacts side by side when analyzing the feared events and consider their severity;
  • to recall the different impacts and their estimation when presenting the risks map (which only keeps the most important value).

This makes it easy to carry out a study of both information security and privacy by presenting side-by-side the impacts to the organization and the impacts to the individuals (data subjects).