PRIVATE ALPHA · INVITE-ONLY DESIGN PARTNERS

Cut SIEM costs ~65% without losing a single log

Trevane runs as a single compiled binary inside your AWS environment between your forwarders and analytical platforms. It aggregates repetitive operational noise into summary records, stores 100% of raw forensic logs in customer-owned S3 Parquet, and delivers sub-15ms SQL queries via local DuckDB.

Run Offline Audit Tool
Invite-only design partner cohort for engineering teams spending $1kโ€“$50k+/mo on Datadog or Splunk ยท Raw logs never leave your VPC. Only summaries go to your SIEM.
71.1%
Volume reduction on a 2.03M-event public dataset stream (Splunk BOTSv3, 1.41 GB)
100%
Raw telemetry retention in S3 Parquet (no NAT or egress fees for S3 traffic)
< 15 ms
Local DuckDB query latency across Parquet shards

System Architecture

Trevane deploys as an in-VPC daemon that intercepts telemetry before network egress or ingestion billing.

1. Telemetry Ingress
AWS CloudTrail (Management)
AWS VPC Flow Logs (Network)
Kubernetes stdout / FluentBit
CrowdStrike Falcon FDR
2. In-VPC Data Plane

Pass-First-N Folding: Baseline heartbeats pass through so absence-of-event monitors stay alive. Traffic spikes automatically unroll into raw stream pass-through so standard COUNT(*) alerts fire without monitor rewrites.

Sub-50ยตs Threat Gate: Mutating IAM and defense evasion actions (StopLogging, DeleteTrail, CreateAccessKey) bypass folding completely with 0% collapse.

Local NVMe Read Cache: Recent Parquet partitions are cached on instance NVMe for sub-15ms on-call deep-link investigation.

3. Persistence Sinks
Tier 1: S3 Parquet Lakehouse
100% Raw Retained ยท No egress fees (VPC gateway endpoint)
Tier 2: Hot SIEM (Datadog/Splunk)
Summary Records + Unrolled Spikes (~65% Cut)

Empirical Benchmarks

Evaluated directly with the compiled Rust kernel (trevane_core) across public security datasets (Splunk BOTSv3, OTRF Mordor, AWS CloudTrail). Measures volume reduction ratios and 100% preservation of security-significant events.

High-Volume Operational Streams (SIEM Ingest & Volume Cut)

Evaluated across sliding 30-second windows. Proves that routine heartbeat, polling, and operational telemetry can be aggregated into summaries while retaining 100% of forensic raw logs in customer-owned S3 Parquet.

Dataset Telemetry Feed Type Events Volume Cut Byte Cut Attack Context
BOTSv3 2.03M Full Stream Capture
1.41 GB multi-source capture (S3, SMTP, Code42, RDS)
Multi-Source Feed 2,036,380 71.1% 68.6% ✓ 100% (3,752 / 3,752)
BOTSv3 100k Stream
Multi-source operational stream capture
CloudTrail + Sysmon + VPC Flow 100,000 71.9% 66.0% ✓ 100% (254 / 254)
BOTSv3 25k Stream Sample
Blended operational feed snapshot
CloudTrail + Sysmon + VPC Flow 25,000 65.3% 63.5% ✓ 100% (60 / 60)
Splunk BOTSv3 Enterprise Suite
Multi-tier attack simulation baseline
Blended CloudTrail + Sysmon 4,559 73.0% 67.6% ✓ 100% (61 / 61)
Splunk BOTSv3 Host Execution
Host process & heartbeat telemetry
Host Endpoint (Sysmon / EDR) 3,616 79.1% 76.9% Baseline (0 attacks)
Splunk BOTSv3 AWS CloudTrail
High-entropy mutating management APIs
AWS Control Plane API 943 53.9% 59.1% ✓ 100% (61 / 61)
Adversarial Attack Invariants (Every Tested Attack Event Forwarded)

When active intrusion payloads or defense-evasion API calls occur, Trevane automatically bypasses rollup folding. 100% of attack payloads are forwarded uncollapsed so downstream Datadog monitors and Splunk correlation searches fire without alert suppression.

Adversarial Scenario MITRE ATT&CK Technique Attack Events Forwarded to SIEM Aggregation Alert Invariant
Stratus Red Team
CloudTrail tampering attack
T1562.008 (StopLogging / DeleteTrail) 3 ✓ 3 / 3 (100%) 0.0% Collapsed Every tested attack event was forwarded
Atomic Red Team
IAM privilege escalation backdoor
T1078 (CreateAccessKey)
67
75 events total
✓ 67 / 67 (100%)
67 of 67 attack events forwarded (75 in total)
0% of attack events collapsed
(1 non-attack event folded)
Every tested attack event was forwarded

Pricing & Private Alpha Cohort

Flat annual software licenses based on daily ingested volume. Zero per-seat surcharges or analytical query fees.

Pilot
For early-stage engineering fleets spending $1kโ€“$5k/mo on Datadog or Splunk
$6,000
$500 / month ยท billed annually
โœ“ Up to 250 GB/day volume
โœ“ Single-node in-VPC forwarder binary
โœ“ S3 Parquet Lakehouse compactor
โœ“ Local DuckDB query console
โœ“ Direct founder onboarding Slack
โœ“ No NAT or egress fees for S3 traffic (VPC gateway endpoint)
Starter
For growing teams spending $5kโ€“$15k/mo on Datadog or Splunk
$30,000
$2,500 / month ยท billed annually
โœ“ Up to 1 TB/day volume
โœ“ Multi-stream (CloudTrail, VPC, K8s)
โœ“ Pass-first-N heartbeat folding
โœ“ Local NVMe SSD query cache
โœ“ Standard engineering Slack support
Scale
For high-throughput multi-account AWS Organizations spending $50k+/mo
$150,000+
Custom deployment license
โœ“ More than 3 TB/day volume
โœ“ Multi-account AWS Organizations sync Planned
โœ“ Custom rules for what must never be folded Planned
โœ“ Help with SOC 2 and HIPAA evidence Planned
โœ“ Priority escalation, response times agreed in contract
Need more volume than your tier? Talk to us and we will move you to the right one.
Private alpha. Items marked Planned are on our roadmap and not yet available.
* Assumes a $500k/yr Datadog or Splunk bill, a 65% volume cut, minus the $75k license and about $10k/yr of AWS infrastructure. Your actual saving depends on your traffic.
Have custom architecture or spend requirements?
We work directly with engineering leadership to scope bespoke pilot deployments and prove exact ROI before production rollout.