Rethinking how host-based intrusion detectors use time
Host-based intrusion detectors order events and attack stages by timestamps but often leave assumptions about clocks and evidence unstated.
Host-based intrusion detectors use timestamps to order events and window audit streams. They also use timestamps to sequence attack stages.
Every conclusion rests on assumptions about clocks, delivery, entity identity, and the integrity of the evidence. Most of these assumptions go unstated.
This Systematization of Knowledge (SoK) makes these assumptions explicit. It distills five components of a time model and a checklist of evidence assumptions.
The work assesses eleven temporal knowledge representation and reasoning (KRR) formalisms. It also assesses a learned baseline on nine attack patterns.
In 28 of 38 audited provenance-era detector papers, captured or stored evidence is trusted. None says which conclusions survive deleted, delayed, or forged records.
Current provenance-based detectors state no clock bound. No paper specifies conditions for conclusion validity across a stated class of delivery failures.
Key facts
- None of the audited papers specify conditions for conclusion validity across a stated class of delivery failures.
- In 28 of 38 audited provenance-era detector papers, the manipulation of captured or stored evidence is placed out of scope.
- Current provenance-based detectors state no clock bound.