Hey everyone, hoping to get some other perspectives on this. I've been running BeyondTrust's Privilege Management for Windows (PMW) in our staging environment for the last three weeks, and while the core privilege elevation blocking works great, I'm consistently noticing what feels like pretty high latency when it intervenes.
Specifically, when a standard user triggers a process that requires elevation and the PMW policy kicks in, there's a noticeable delay—sometimes 2 to 4 seconds—before the process actually starts or the user gets the "denied" or "elevated with credentials" prompt. This is happening even on relatively fresh, well-spec'd VMs. It's not a deal-breaker, but it does impact the user experience, making things feel sluggish.
Here’s what I’ve checked so far in our setup:
* **Agent Version:** We're on PMW 23.5.
* **Policy Complexity:** Our test policy isn't huge—maybe 50 rules right now, mixing file path, hash, and publisher conditions.
* **Management Server:** The agent is communicating with a dedicated server on the same local network, ping times are <1ms.
* **Logging:** I have debug logging enabled on a test machine. The logs show the delay seems to be between the `ProcessCreate` event and the policy evaluation result.
A snippet from the local agent log with timestamps tells the story:
```
2024-05-15 14:32:01.123 INFO ProcessCreate: C:UserstestAppDataLocalTempinstaller_tool.exe
2024-05-15 14:32:03.845 INFO Policy evaluation complete. Action: ElevateWithCredentials. Rule: "Allow signed vendors".
```
That's over 2.5 seconds just for evaluation! 😅
I'm curious if others are seeing similar behavior. Could this be related to:
* The order or type of conditions in the rules? (e.g., checking a file hash first vs. a publisher certificate)
* Having the management server online vs. a cached/local policy mode?
* Something specific with the 23.5 agent?
I plan to do some more isolated tests this week—like stripping the policy down to just one rule and testing with the server offline—but wanted to see if this rings any bells or if my expectations for near-instantaneous response are off. Any insights or your own latency observations would be super helpful!
Yeah, latency's the hidden tax for these tools. They all have it. You said your policy is "maybe 50 rules." That's not huge to you, but it's a decent amount for the agent to parse and evaluate in real-time on every process launch. Even on an SSD.
The 2-4 second delay is probably the agent checking in with your management server for a policy update or decision, despite the low ping. Is real-time reporting or enhanced logging turned on? That'll add to the weight.
Check if your rules have a lot of "file hash" conditions. Calculating that on the fly for every .exe adds overhead. It's the price of the "trust nothing" model, I guess. 😏
Buyer beware