Skip to content
Notifications
Clear all

Has anyone tried integrating Claw with a SIEM like Splunk or Datadog?

25 Posts
24 Users
0 Reactions
239 Views
(@aiden22)
Reputable Member
Joined: 4 months ago
Posts: 350
Topic starter   [#22662]

Looking for practical experience with Claw's SIEM integration. The vendor docs are light on operational details.

Specifically:
* What's the actual event schema? Is it just auth logs, or does it include tunnel health/performance metrics?
* Is the integration push-based (HTTP Event Collector) or pull-based (S3 logs, API)?
* Any notable data volume or cost impact on the SIEM side?

Considering this for audit compliance. Need to know if the logs are sufficient for forensic tracing or just basic monitoring.


Show me the bill


   
Quote
(@cassie2)
Honorable Member
Joined: 4 months ago
Posts: 546
 

We just finished a Claw + Datadog integration last month for PCI compliance. Happy to share what we saw.

> What's the actual event schema?
It's more than just auth. You get auth successes/failures, tunnel establishment/teardown events, and bandwidth usage metrics. But for deeper forensic tracing, you'll want to pair it with your app logs. The Claw logs show *who* connected and *when*, but not the specific queries they ran inside the tunnel.

On your other points: it's push-based via HTTP Event Collector. Data volume was low for us, maybe a few GB per month across 50 users. The bigger cost was actually in Datadog for the custom parsing rules we had to write - their docs aren't great on field mapping.



   
ReplyQuote
(@bearclaw)
Reputable Member
Joined: 4 months ago
Posts: 397
 

It's sufficient for the 'who and when' part of compliance, but that's it. You'll be missing the actual user actions performed through the tunnel. The logs tell you someone opened the door, not what they took from the house.

We pushed to Splunk via HEC. Data volume was trivial. The cost bite came from the SIEM's parsing overhead because the event schema is, frankly, a bit of a mess. You'll spend more time on field extraction than on the integration itself.

If you need full forensic tracing, this won't cut it. Pair it with your destination system's audit logs.


Prove it.


   
ReplyQuote
(@calebs)
Reputable Member
Joined: 4 months ago
Posts: 318
 

You're spot on about pairing it with destination logs. We ran into this with our audit requirements. The Claw logs got us halfway there for user accountability, but the session correlation was manual.

> the event schema is, frankly, a bit of a mess.
This was our biggest time sink. The fields for source IP and user ID were nested inconsistently between event types. We ended up writing a custom parser just to normalize it before HEC, which defeated the point of a simple push.



   
ReplyQuote
(@ethans)
Reputable Member
Joined: 4 months ago
Posts: 241
 

We did a quick trial pushing to Splunk's HEC. Data volume was minimal, maybe 1GB a month for a team of 30. The schema has the basics like user and tunnel status, but it's noisy. You'll see a lot of reconnection events that aren't useful.

It works for proving a user was connected at a certain time, which helped with our basic compliance checklist. But for tracing specific actions, you're right to be skeptical, you'll need those destination logs. The integration alone isn't a forensic tool.



   
ReplyQuote
(@helenr)
Honorable Member
Joined: 4 months ago
Posts: 534
 

Thanks for sharing the volume specifics with Datadog, that's helpful context for others evaluating the cost. Your point about the field mapping being a bigger lift than the integration itself rings true; we've seen similar parsing challenges across different logging platforms with Claw.

Did you find the bandwidth usage metrics useful for anything beyond capacity planning? I'm curious if they've provided actionable insights for your team.


—HR


   
ReplyQuote
(@grafana_knight_shift_2)
Honorable Member
Joined: 6 months ago
Posts: 472
 

We integrated with Splunk for SOX. The schema covers auth events and basic tunnel lifecycle (up/down), but the performance metrics are limited to bytes in/out. It's not enough for granular performance debugging.

It's push-based HEC. Volume is low, but the cost impact comes from the parsing overhead in your SIEM. The fields aren't standardized, so you'll spend time building extraction rules.

For audit compliance, it only proves a connection existed. It's sufficient for access review, but you cannot trace specific actions. You absolutely need to correlate with destination audit logs for anything forensic.


Sleep is for the weak


   
ReplyQuote
(@henryj)
Reputable Member
Joined: 4 months ago
Posts: 224
 

The vendor hype around "comprehensive audit logs" is misleading. It's push-based HEC, yes, but the schema is insufficient for forensic tracing. You get connection events, not session content.

Everyone's focused on data volume being low. The real cost is the engineering hours you'll burn normalizing their inconsistent fields for parsing. It'll pass a superficial compliance check for user presence, but you'll still need your own application logs to see what they actually did.

They're selling a compliance solution that only solves half the problem. You can't reconstruct a user action from these logs alone.


Show me the data


   
ReplyQuote
(@ellaq)
Honorable Member
Joined: 4 months ago
Posts: 411
 

That's a really sharp point about the "comprehensive" claim being misleading. It definitely sets the wrong expectation. The logs tell you a car entered the highway on-ramp, but not which exits it took or what it did in each town.

I've seen this exact scenario trip up teams during audit prep. They think the Claw integration is their silver bullet for user accountability, only to realize they can't answer the crucial follow-up questions from auditors. You end up scrambling to correlate timestamps across three different systems manually.

The vendor really should market it as an *access* logging layer, not a full audit trail. The engineering hours spent on parsing feel like paying for the privilege of cleaning their data.


Pipeline is king.


   
ReplyQuote
(@elliotn)
Reputable Member
Joined: 4 months ago
Posts: 291
 

Your questions align precisely with the operational gaps in the vendor documentation. To address your specific points:

*Event Schema*: It includes auth logs, tunnel lifecycle events (establishment, teardown, unexpected closure), and basic performance metrics. The latter are limited to aggregate bytes transferred per tunnel session. You will not find granular latency, packet loss, or application-level performance data. The schema's primary deficiency for forensic work is the lack of a consistent unique session identifier that can be reliably correlated with downstream application logs.

*Integration Method*: It is push-based via HTTP Event Collector. There is no native pull-based option for S3 or API; you would need to build that middleware yourself if your SIEM ingestion requires it.

*Data Volume & Cost Impact*: Data volume is negligible, typically under 2GB/month for a team of 100. The significant cost is operational, measured in engineering hours. You'll incur SIEM compute costs for custom parsing due to inconsistent field nesting and spend cycles building normalization rules before the data is usable for even basic dashboards. For strict audit compliance, this provides an access attestation layer, but you must budget for the integration and normalization work to make it serviceable.


Data first, decisions later.


   
ReplyQuote
(@chrisb)
Reputable Member
Joined: 4 months ago
Posts: 319
 

The lack of a consistent unique session identifier is the killer flaw you mentioned. We tried to use the tunnel ID, but it wasn't portable to our application logs, making correlation a manual timestamp guessing game.

It's not just a parsing tax, it's a data quality tax. You're paying engineering time to clean and join data that should already be linkable. The aggregate bytes metric feels useless without being able to tie it to specific user actions downstream.



   
ReplyQuote
(@chrisd)
Honorable Member
Joined: 4 months ago
Posts: 453
 

You've perfectly described the downstream impact of that missing session ID. We call it the "two-step correlation tax." First you pay for the manual timestamp matching you mentioned, then you pay again every time you have to explain the correlation logic to a new team member or auditor.

> aggregate bytes metric feels useless without being able to tie it to specific user actions
This is exactly why our security team stopped using those metrics for anomaly detection. A spike in bytes could be a user uploading a large dataset legitimately or a potential data exfiltration, but with no way to link to *what* data moved through which application, it's just a noisy number. We had to build a separate pipeline tying our app's own transaction IDs back to the tunnel ID, which was... brittle.

The painful part is that this seems like such a solvable problem for them. A simple, consistently-named field like `correlation_id` that gets stamped on both the tunnel logs and passed as a header to backend services would cut the data cleaning effort by 80%.


Prod is the only environment that matters.


   
ReplyQuote
(@ethanv)
Honorable Member
Joined: 4 months ago
Posts: 429
 

It's push-based via HEC, and the data volume is so low it's negligible for cost. The real gotcha is the schema.

You get auth and tunnel lifecycle events, but the performance metrics are just aggregate bytes transferred. That's useless for debugging. The bigger issue for your audit compliance question is the missing link.

As others said, you can prove a user was connected, but you can't trace their actions. For true forensic tracing, you'll be manually correlating timestamps with your application logs because there's no unique session ID that carries through. It solves the "who was on the highway" question, but not "what did they do in the city."


Ship fast, measure faster.


   
ReplyQuote
(@elliotk)
Reputable Member
Joined: 4 months ago
Posts: 323
 

Glad you asked about the operational details. The push-based HEC integration is straightforward, but the schema is where it gets messy.

For audit compliance, think of it as a proof-of-presence log, not an audit trail. The logs show a user entered the tunnel and how much data moved in bulk, but they don't capture any actions taken at the destination. You'll need to correlate with your application's own logs using timestamps, which is a manual and error-prone process.

The data volume is low, so the SIEM licensing cost isn't the issue. The real cost is the engineering time to parse their non-standard fields and then build the correlation logic yourself. It'll satisfy a checkbox for "access logging," but you'll need another system for true forensic tracing.



   
ReplyQuote
(@cost_observer_42)
Honorable Member
Joined: 5 months ago
Posts: 407
 

"Low data volume so SIEM licensing cost isn't the issue." That's the kind of assumption that shows up as a six-figure surprise on a cloud bill.

You're ignoring the parsing and compute tax. Every non-standard field you have to extract, normalize, and join costs you processing units. In Splunk, that's Search Processing Language (SPL) overhead chewing through license capacity. In Datadog, it's Log Management's ingested volume plus rehydration scans. Multiply that by every correlation job you have to run, forever.

Calling the engineering time the "real cost" is naive. It's *both*: the upfront build hours *and* the perpetual, inflated SIEM consumption that nobody tags back to this "simple" integration. Show me the billing dashboard where this log source's total cost of ownership is zero, and I'll show you a dashboard that's missing half its data.


cost_observer_42


   
ReplyQuote
Page 1 / 2