Case Study Telecommunications

Detecting a supply chain compromise 6 weeks before impact

A Tier-1 European telco identified anomalous authentication behavior in a third-party integration 6 weeks before it would have caused a service disruption — a signal buried in logs their SIEM had been configured to drop.

6 weeks

Earlier detection vs estimated impact date

100%

API and integration log coverage

0

Customer-impacting incidents in 12 months post-deployment

1000×

Inference cost reduction

€2.8M

Estimated avoided remediation cost

11 days

Time to detect anomalous integration behavior

The Organisation

European Telecommunications Group · Telecommunications

The Challenge

The telco operated infrastructure across 12 countries with thousands of third-party integrations. Their security team had grown accustomed to working with sampled telemetry — a pragmatic response to the cost of full ingestion. The specific logs being dropped were those generated by low-priority service accounts and API integrations — exactly the category attackers had begun targeting in supply chain campaigns.

The Approach

Full coverage LLM reasoning was applied to all API access logs, service account activity, and integration telemetry. The LLM model was tasked with identifying behavioral anomalies — not just rule violations — across the complete event stream.

"The pattern was subtle — a service account authenticating at unusual hours with slightly elevated API call rates. Nothing that would have triggered a rule. The LLM flagged it as anomalous because it had the full behavioral context. Our old system never would have seen it."

Head of Security Operations

Key Finding

The compromised integration belonged to a third-party monitoring vendor. The attacker had established a low-and-slow foothold using legitimate credentials, incrementally escalating access over several weeks. At sampled telemetry coverage the activity was indistinguishable from normal operation. At full coverage, the cumulative behavioral signature was clear within 11 days of the initial compromise.

Results at a Glance
Earlier detection vs estimated impact date 6 weeks
API and integration log coverage 100%
Customer-impacting incidents in 12 months post-deployment 0
Inference cost reduction 1000×
Estimated avoided remediation cost €2.8M
Time to detect anomalous integration behavior 11 days
Get in Touch

Talk to us about your data.

Tell us about your event stream and we'll show you what full LLM reasoning coverage looks like for your environment.

Or book a call directly →