Previous risk calculation results not fully cleared

[mike1813]

When the risk calculator runs, there is a call to RiskCalculator.initialiseLevels(), which does two things:

• it sets initial values for every likelihood/TW level that is to be calculated, and
• it clears threat graph properties indicating which TWA are external causes, which threats are initial or root causes, which threats cause which misbehaviours, and so forth

Unfortunately, the second step is incomplete because some threat graph properties have been added since RiskCalculator.initialiseLevels() was last updated.

The one I noticed was the property core#isNormalOpEffect of MisbehaviourSet entities, which is set to true if the associated behaviour is caused by normal operational threats. This doesn't sound like a risk calculation output because it seems like it should be a constant property of the behaviour. In fact, it is possible for behaviours to be caused by both normal operational and adverse threats, and in that case which threat is considered responsible depends on which threat is more likely - i.e., the result of the risk calculation given the assumed TW levels and control coverage levels.

The way to spot this is to validate a model, then run three risk calculations, exporting the resulting NQ files:

• a CURRENT risk calculation
• a FUTURE risk calculation
• a CURRENT risk calculation

The FUTURE risk calculation considers a lot of extra threats, many being normal operational threats, e.g., threats modelling the discovery of softwaere vulnerabilities. The core#isNormalOpEffect property gets set for various behaviours in the second risk calculation in this sequence. However, those threats are ignored in a CURRENT risk calculation so the core#isNormalOpEffect property should not be set in the first and last calculations.

It turns out that this property is absent (as it should be) in the results of the first CURRENT risk calculation, but not the second. This is because it gets added by the FUTURE risk calculation, and is not cleared properly by RiskCalculator.initialiseLevels() in the subsequent CURRENT risk calculation.

[mike1813]

Reviewed by @mike1813 : this is a bug that is likely to affect many users.

It may be related to issue #50, although that is concerned with the retention of volatile output for use in a subsequent calculation, where this issue is about negating previous output before a new calculation. Unlike #50, it should be an easy fix.