I'm not happy with the idea of 'levels' in many contexts, including information classification schemes. The term 'level' implies a stepped progression in one dimension. Information risk and security is more nuanced or fine-grained than that, and multidimensional too.
The problems with 'levels' include:
- Boundary/borderline cases, when decisions about which level is appropriate are arbitrary but the implications can be significant;
- Dynamics - something that is a medium level right now may turn into a high or a low at some future point, perhaps when certain event occurs;
- Context e.g. determining the sensitivity of information for deliberate internal distribution is not the same as for unauthorized access, especially external leakage and legal discovery (think: internal email);
- Dependencies and linkages e.g. an individual data point has more value as part of a time sequence or data set ...
- ... and aggregation e.g. a structured and systematic compilation of public information aggregated from various sources can be sensitive;
- Differing perspectives, biases and prejudices, plus limited knowledge, misunderstandings, plain mistakes and secret agendas of those who classify stuff, almost inevitably bringing an element of subjectivity to the process despite the appearance of objectivity;
- And the implicit "We've classified it and [maybe] done something about securing it ... so we're done here. Next!". It's dismissive.
The complexities are pretty obvious if you think about it, especially if you have been through the pain of developing and implementing a practical classification scheme. Take a blood pressure reading, for instance, or an annual report or a system security log. How would you classify them? Whatever your answer, I'm sure I can think of situations where those classifications are inappropriate. We might agree on the classification for a particular situation, hence a specific level or label might be appropriate right there, but information and situations are constantly changing, in general, hence in the real world the classification can be misleading and unhelpful. And if you insist on narrowing the classification criteria, we're moving away from the main advantage of classification which is to apply broadly similar risk treatments to each level. Ultimately, every item needs its own unique classification, so why bother?
Another issue with classification schemes is that they over-emphasize one aspect or feature of information - almost always that's confidentiality. What about integrity, availability, utility, value and so forth? I prefer a conceptually different approach using several tags or parameters rather than single classification 'levels'. A given item of information, or perhaps a collection of related items, might usefully be measured and tagged according to several parameters such as:
- Sensitivity, confidentiality or privacy expectations;
- Source e.g. was it generated internally, found on the web, or supplied by a third party?;
- Trustworthiness, credibility and authenticity - could it have been faked?;
- Accuracy and precision which matters for some applications, quite a lot really;
- Criticality for the business, safety, stakeholders, the world ...;
- Timeliness or freshness, age and history, hinting at the information lifecycle;
- Extent of distribution, whether known and authorized or not;
- Utility and value to various parties - not just the current or authorized possessors;
- Probability and impact of various incidents i.e. the information risks;
The tags or parameters required depend on what needs to be done. If we're determining access rights, for instance, access-related tags are more relevant than the others. If we're worried about fraud and deception, those integrity aspects are of interest. In other words, there's no need to attempt to fully assess and tag or measure everything, right now: a more pragmatic approach (measuring and tagging whatever is needed for the job in hand) works fine.
Within each parameter, you might consider the different tags or labels to represent levels but I'm more concerned with the broader concept of taking into account a number of relevant parameters in parallel, not just sensitivity or whatever.
All that complexity can be hidden within Gary's Little World, handled internally within the information risk and security function and related colleagues. Beyond that in the wider organization, things get messy in practice but, generally speaking, people working routinely with information "just know" how important/valuable it is, what's important about it, and so on. They may express it in all sorts of ways (not just in words!), and that's fine. They may need a little guidance here and there but I'm not keen on classification as a method for managing information risk. It's too crude for me, except perhaps as a basic starting point. More useful is the process of getting people to think about this stuff and do whatever is appropriate under the circumstances. It's one of those situations where the journey is more valuable than the destination. The analysis generates understanding and insight which are more important that the 'level'.