Oracle SBC Security Guide
Per-endpoint Call Admission Control
The SBC can demote endpoints from trusted to untrusted, or untrusted to denied queues when CAC
failures exceed a configured threshold. The SBC maintains CAC failures per-endpoint. The CAC failure
counter is incremented upon certain admission control failures only if either: cac-failure-threshold or
untrust-cac-fail-threshold is set to a non-zero integer.
The cac-failure-threshold parameter is configurable in the access control and realm configuration
elements. Exceeding the threshold integer defined in this parameter demotes an endpoint from the trusted
queue to the untrusted queue. Additionally, the untrust-cac-failure-threshold parameter is configurable in
the access control and realm configuration elements. Exceeding the threshold integer defined in this
parameter demotes an endpoint from the untrusted queue to the denied queue. If both the cac-failure-
threshold and untrust-cac-failure-threshold are configured to 0, admission control failures are considered
and counted as invalid signaling messages for determining if the invalid-signal-threshold parameter value
has been exceeded.
CAC failures used for Endpoint Demotion
The SBC determines CAC failures only by considering the number of signaling messages sent FROM an
endpoint TO the realm its signaling messages traverse.
If an endpoint exceeds the following CAC thresholds, the SBC will demote the endpoint when the CAC
failure thresholds are enabled.
sip-interface user CAC sessions (realm-config > user-cac-sessions)
sip-interface user CAC bandwidth (realm-config > user-cac-bandwidth)
External policy server rejects a session
Thresholds and Trending Analysis
Thresholds and trending analysis are important concepts that must be well understood and implemented
during initial installation of the SBC. Thresholds should be monitored and settings periodically adjusted
as network usage or capacity requirements change. To be supported by Oracle TAC, SBC deployments
require a minimum set of standard configurations outlined in the DDoS BCPs [10, 11]. These settings are
considered the minimum configuration required to protect the SD. Upon deployment of a DDoS
provisioned SBC it’s recommended that customers continuously monitor common traffic load and
patterns of services traversing their SBC, and understand any alarms received.
Regardless of the monitoring method used (i.e. SNMP, CDR, HDR, Syslogs), during the initial period
after implementation it’s crucial to understand:
The number of active SIP sessions seen during normal and peak periods
Average call hold times
Average signaling messages for a call (usually best collected via Wireshark or other network
capture tool)
What are the stable and “common” values of these for the different counters
o Trusted to Untrusted Demotions
o Untrusted to Deny Demotions
o Demotions
o Promotions
On-going demotions/promotions on ACLs, and to which SIP UAs they refer to
Why there are any deny entries and to which SIP UAs they refer to
Whether the deny period set is helping or causing more issues
Kommentare zu diesen Handbüchern