RPKI cover hijack correlation logic¶
What this correlation detects¶
An active BGP hijack or interception in which an adversary:
Announces a prefix or sub-prefix
Ensures the announcement appears RPKI-valid
Achieves routing acceptance under that cover
Operates briefly, then withdraws or stabilises
Unlike noisy hijacks, this attack succeeds by abusing trust, not ignoring it.
Why this matters¶
Classic hijack detection looks for:
invalid routes
sudden origin changes
obvious leaks
This attack does none of that.
If you only alert on invalid, you miss it.
If you alert on every announcement, you drown.
Only correlation across systems reveals intent.
Observable signals¶
Observable without payload inspection:
BGP announcements and withdrawals for monitored prefixes (via BMP)
Route attributes consistent with legitimate origin
RPKI validation results indicating
validPrior or concurrent ROA state changes (contextual, not required)
No DPI. No traffic capture. No heroics.
Required log sources¶
This correlation assumes access to:
BMP (BGP Monitoring Protocol): Authoritative view of announcements and withdrawals
RPKI validator logs: Validation outcome for the observed prefix–origin pairs
Optional but useful:
Historical ROA change records
Prefix ownership metadata
Router syslog is not required and not authoritative for routing truth.
Correlation logic (human-readable)¶
The detection logic is as follows:
A BGP announcement for a monitored prefix or sub-prefix is observed via BMP
The announcement is classified as RPKI-
validby one or more validatorsThe route attracts traffic (inferred by propagation or persistence)
The prefix is later withdrawn or replaced without operational justification
When a routing event both looks legitimate and behaves tactically, raise a high-confidence alert for an RPKI-covered hijack.
Temporal considerations¶
Announcement and validation occur close in time
Withdrawal may be delayed
Correlation must allow non-symmetric timing
Short attack, long memory.
Assumptions and limitations¶
Assumptions:
BMP visibility exists for relevant peers
Validator results are trustworthy enough for consensus
Limitations:
Persistent hijacks may evade withdrawal-based logic
Partial BMP visibility reduces confidence
Detection is late-stage by design
This favours accuracy over immediacy.
Evasion considerations¶
A capable adversary may:
Maintain the route indefinitely
Announce exact prefixes only
Limit propagation to reduce visibility
Each evasion trades stealth for operational complexity.
Expected outcome¶
A triggered correlation indicates:
Routing trust has been successfully abused
Immediate mitigation is required
Trust anchors and ROAs must be audited
This is not a false positive you ignore. This is a control-plane incident.