Combining Generated Data Models with Formal Invalidation for Insider Threat Analysis

In this paper we revisit the advances made on invalidation policies to explore attack possibilities in organizational models. One aspect that has so far eloped systematic analysis of insider threat is the integration of data into attack scenarios and its exploitation for analyzing the models. We draw from recent insights into generation of insider data to complement a logic based mechanical approach.

We show how insider analysis can be traced back to the early days of security verification and the Lowe-attack on NSPK. The invalidation of policies allows modelchecking organizational structures to detect insider attacks. Integration of higher order logic specification techniques allows the use of data refinement to explore attack possibilities beyond the initial system specification.

We illustrate this combined invalidation technique on the classical example of the naughty lottery fairy. Data generation techniques support the automatic generation of insider attack data for research. The data generation is however always based on human generated insider attack scenarios that have to be designed based on domain knowledge of counter-intelligence experts.

Introducing data refinement and invalidation techniques here allows the systematic exploration of such scenarios and exploit data centric views into insider threat analysis.