Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Distributed tracing follows a single request across every service it touches. For security teams it turns a scattered set of logs into one reconstructable chain of events.

What Is Distributed Tracing in Security?
Distributed tracing records the journey of a single request through a system made of many services. Each unit of work becomes a span, carrying a start time, a duration, a parent, and a set of attributes; the spans sharing one trace identifier reconstruct the whole path. It was built for performance engineering—finding which of fourteen microservices caused a slow checkout—but it answers a question security teams ask constantly and usually cannot: what exactly happened, in order, when that request arrived.
In a microservice architecture, a single user action produces log lines in a dozen places with no shared identifier. Correlating them means guessing from timestamps, which fails under load and fails completely when retries, queues, and asynchronous workers are involved. An investigator trying to establish whether a specific request reached the database, which identity it ran under by the time it got there, and what it returned may spend hours on a question tracing answers in seconds.
OpenTelemetry, now the de facto standard under the Cloud Native Computing Foundation, propagates context between services using the W3C Trace Context headers traceparent and tracestate. Jaeger, Grafana Tempo, AWS X-Ray, Azure Application Insights, Google Cloud Trace, and the commercial observability platforms all consume it.
Spans attract attributes, and developers add whatever is convenient for debugging. That routinely includes full URLs with query strings, request and response bodies, authentication headers, session tokens, customer identifiers, and occasionally the contents of a form. The trace backend is then shared widely across engineering, retained for weeks, and frequently hosted by a third party.
Treat tracing data as production data. Define an attribute policy stating what may be recorded; strip credentials and tokens in the instrumentation rather than in the backend; hash or tokenise identifiers that only need to be correlatable; apply redaction at the collector as a second line; restrict who can query the backend; and include the trace platform in your records of processing under UK GDPR, because it almost certainly holds personal data.
Trace headers arriving from outside your perimeter are user input. An attacker can supply a traceparent to join their request to an existing trace, inject misleading parent relationships, or stuff oversized values into tracestate to disrupt processing. Decide deliberately whether external trace context is accepted or whether the edge always starts a new trace and records the submitted value as an attribute. Treat trace identifiers as correlation aids, never as evidence of authenticity, and never make an authorisation decision based on one.
Sampling is the practical obstacle. Performance teams sample heavily—often one per cent or less—and the trace you want during an investigation is exactly the one that was discarded. Use tail-based sampling that retains all errors, all unusually slow requests, and all traces touching sensitive endpoints, and keep head-based sampling for the ordinary majority.
Then connect it to the rest of your telemetry. Emit the trace identifier into application logs so a log line links to its trace; include it in audit records; and carry it into alerts so a responder lands on the full chain rather than a single event. Done well, tracing becomes the connective tissue between the application layer and the control-plane telemetry described in cloud logging and monitoring.
At Privia Security, we help teams turn observability data into usable detection through our SOC maturation services.
Not primarily, and it does not replace logging, audit trails, or detection tooling. It is an observability capability that happens to answer questions security teams struggle with, particularly in microservice architectures where a single request touches many services. Treat it as a high-value investigation aid and a source of behavioural signal, built and owned by engineering, with security as a consumer.
Most often full URLs including query parameters, request and response bodies, authorisation headers and session tokens, customer identifiers, email addresses, and occasionally payment or health data copied in during debugging. Define an attribute policy, redact at the point of instrumentation, add collector-side redaction as a backstop, and audit a sample of spans regularly to confirm it is working.
Yes. Trace context arriving from the internet is attacker-controlled input and can be used to join a request to an existing trace, forge parent relationships, or submit oversized values. Decide whether your edge accepts external trace context at all, record rather than trust any submitted value, and never use a trace identifier in an authentication or authorisation decision.
Severely, if it is left at performance-engineering defaults. Sampling one per cent of traces means the request you need during an incident is almost certainly gone. Use tail-based sampling so that errors, unusually slow requests, and anything touching authentication or sensitive endpoints are always retained, and keep aggressive sampling only for routine successful traffic.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.