Definition
Percentage of known isolation threat categories (traffic, compute, storage, timing/side-channel, metadata/correlation) for which the monitoring system has demonstrated detection capability, measuring whether isolation assurance covers the full threat surface or only the obvious layers.
The Measurement Gap
Measurement Gap and Decision
Decision Supported: Is the isolation guarantee enterprise customers are paying for real, or is it only enforced at the network layer while side-channel, timing, and metadata threats go unmonitored?
Measurement Gap:
Enterprise customers purchasing dedicated network slices pay a premium for isolation. The guarantee that their traffic, compute, and data are separated from other tenants. Operators enforce and monitor this isolation at the network layer: VLANs, VRFs, dedicated bearers. But isolation threats exist at multiple layers: compute resource interference, storage I/O contention, timing side-channels that leak information between co-located workloads, cache side-channels, and metadata correlation attacks that infer tenant behavior from observable network patterns. Most operators monitor one or two of these layers. None systematically assess their detection coverage across the full threat taxonomy. An operator can honestly claim "we monitor isolation" while being blind to the majority of isolation threats. IADS quantifies this coverage gap by defining the full threat taxonomy and measuring which categories have demonstrated (not theoretical) detection capability, validated through red-team testing.
Citations:
- ETSI GS NFV-SEC 012, "NFV Security: System Architecture and Specification for Network Slicing," 2024. Defines isolation requirements without measuring detection coverage across the full threat surface.
- 3GPP TR 33.811, "Study on Security Aspects of 5G Network Slicing," Release 18: identifies side-channel and metadata threats without proposing an operational coverage metric.
Formula
Plain Language: Define the full taxonomy of cross-slice isolation threats. For each category, assess whether the monitoring system has demonstrated detection capability (through testing, not just theoretical coverage). Count the categories with demonstrated capability.
Notation:
IADS = Categories_with_demonstrated_detection / Total_isolation_threat_categories × 100
Variable Definitions:
- Categories_with_demonstrated_detection = isolation threat categories where detection has been validated through red-team testing or controlled experiments (count)
- Total_isolation_threat_categories = all known categories of cross-slice isolation threats (count)
Standard taxonomy (proposed starting point): network traffic leakage, compute resource interference, storage I/O interference, timing side-channel, cache side-channel, metadata correlation, API cross-contamination. 7 categories total
Units: Percentage. 100% = all known threat categories have demonstrated detection. Lower values indicate isolation is assured on some layers but not others.
Sign Convention:
Higher is better. 100% is comprehensive isolation assurance.
Edge Cases:
- New threat categories emerge: add to the taxonomy and re-assess. IADS will drop when new threats are added. This is correct behavior.
- "Demonstrated detection" requires an actual test, not a vendor claim. Schedule annual red-team exercises per category.
- Some categories may be infeasible to monitor in real-time (certain side-channel attacks): count as unmonitored and note the limitation.
- This is an assessment metric, not a continuous measurement. Update quarterly or after any architecture change.
Proposed formula, pending validation pilots.