I started testing larger values here by increasing this to 5,000 bytes. That's a ten fold increase of the default, but still far smaller than the recommended value of 65,535. The change was startling, as we saw an immediate increase in the number of alerts, some from rules that had never fired before. I monitored two of the busiest sensors in the system and saw no noticeable hit in performance.
To cut to the chase, I tried values of 10,000, 50,0000 and finally the recommended 65,535 bytes. None of those values gave me an unacceptable performance hit on the sensors, but each time the volume of alerts, what had been false negatives, went up in large measure.
The amount of tuning needing done on the new, heretofore unseen traffic was on par with having added a new segment to monitor. It was amazing and disconcerting how much traffic that low default setting had blinded the sensors to.
The moral of the story here is check every configuration item carefully and make sure you understand what each one does. IDS is a complex beast and you might be missing a lot traffic you should be seeing if you're not careful.
1 comment:
Hi Jeff,
Thanks for the very helpful post. I was thinking of an optimum value and higher values hit performance bad, since number number of searches increase. How is it if we use an inspect_uri_only with higher value of server_flow_depth? Will that be a fair trade off?
Thanks
Post a Comment