Know every vulnerabilitybefore it knows you.
DevGuard continuously monitors your dependencies and alerts you when CVEs like this one affect your stack — with real-time threat intelligence built for developers.
GHSA-4jvg-4jfx-fmhc
No affected components available
Summary
The Sentry exporter constructs Sentry API URLs by interpolating the span's service.name resource attribute into the URL path without validation. Because service.name is controlled by remote OTLP senders and the operator-configured bearer token is attached to every request, a crafted service name can reach arbitrary Sentry API endpoints reachable by that token — including privileged admin, organization, and member endpoints within the configured Sentry organization.
Affected
- exporter/sentryexporter/sentry_exporter.go (lines 715–737) — extractProjectSlug returns the attacker-controlled service.name directly as the slug.
- exporter/sentryexporter/sentry_exporter.go (lines 745–809) — getOrCreateProjectEndpoint passes the raw slug to GetOTLPEndpoints at line 761.
- exporter/sentryexporter/sentry_client.go (lines 190–244) — GetProjectKeys interpolates the slug into fmt.Sprintf URL path and attaches the operator bearer token on line 207.
- exporter/sentryexporter/sentry_client.go (lines 327–363) — GetOTLPEndpoints calls GetProjectKeys on line 329 with the raw slug.
- exporter/sentryexporter/config.go (lines 55–108) — projectSlugRegexp is applied only to operator config mappings inside validateRoutingConfig, never to
runtime-derived slugs.
Root cause
- extractProjectSlug (sentry_exporter.go:715–737) reads service.name from pcommon.Resource.Attributes() without schema validation and returns the raw
string on line 736. - GetProjectKeys (sentry_client.go:192) calls fmt.Sprintf("%s/api/0/projects/%s/%s/keys/", c.baseURL, orgSlug, projectSlug). The slug is treated as a single path segment but no validation is performed.
- projectSlugRegexp (config.go:58) — defined as ^[a-z0-9_-]{1,50}$ — is referenced only inside validateRoutingConfig on line 98 (config-time only). No runtime callsite exists.
- Go net/http preserves literal .. and / characters in URL paths when constructed via fmt.Sprintf.
- The operator-configured DSN / bearer token is attached unconditionally to every outbound request (sentry_client.go:207): req.Header.Set("Authorization", fmt.Sprintf("Bearer %s", c.authToken)).
Exploitation
Primary: query-string injection (reliable across all deployments)
Attacker emits service.name = "foo?injected_query=".
URL becomes https://sentry.io/api/0/projects/ORG-SLUG/foo?injected_query=/keys/.
The trailing /keys/ is consumed as part of the query string. The resource endpoint is /api/0/projects/ORG-SLUG/foo. The attacker can reach any GET-based
Sentry API endpoint reachable by the bearer token. This vector is not dependent on server-side path normalization and works in all deployment
configurations.
Secondary: path traversal (nginx-dependent)
Attacker emits a span with service.name = "foo/../../members".
Resulting URL: https://sentry.io/api/0/projects/ORG-SLUG/foo/../../members/keys/
After server-side normalization (nginx resolves .. segments): https://sentry.io/api/0/projects/ORG-SLUG/members/keys/
The operator bearer token authenticates the request. Effectiveness depends on whether the Sentry deployment normalizes .. segments before routing (standard nginx behaviour).
Amplified: telemetry redirect for data exfiltration
Attacker-owned Sentry project slug → span data for other applications is exported to an attacker-controlled Sentry project, leaking operational telemetry.
The collector fetches the DSN/keys for the attacker's slug and subsequently forwards legitimate traces/logs to the attacker-controlled destination.
Threat model
- Attacker capabilities: remote OTLP trace sender (application-level span emission).
- Operator capabilities: configures Sentry DSN, bearer token, base URL; sets up receiver pipeline.
- The attacker does NOT control operator YAML. The attacker DOES control resource attribute values on spans they emit.
Realistic deployment
- Kubernetes cluster with OpenTelemetry Collector forwarding traces from multiple applications to Sentry SaaS or self-hosted Sentry.
- One compromised or malicious application reaches the collector via OTLP.
- The collector is configured with a valid Sentry bearer token for the organization.
Remediation
Apply the existing projectSlugRegexp to runtime-derived slugs, not only to operator config mappings:
import "regexp"
var runtimeSlugPattern = regexp.MustCompile(`^[a-zA-Z0-9_-]+$`)
func (s *endpointState) extractProjectSlug(attrs pcommon.Map) string {
attrValue, exists := attrs.Get(s.attributeKey)
if !exists || attrValue.Type() != pcommon.ValueTypeStr {
return ""
}
serviceName := attrValue.Str()
if serviceName == "" {
return ""
}
if s.projectMapping != nil {
if mappedSlug, ok := s.projectMapping[serviceName]; ok {
return mappedSlug
}
}
if !runtimeSlugPattern.MatchString(serviceName) {
return "" // reject; drop the span or use a fallback default project
}
return serviceName
}
Alternatively, reject at URL construction:
func (c *sentryClient) GetProjectKeys(ctx context.Context, orgSlug, projectSlug string) ([]projectKey, error) {
if !runtimeSlugPattern.MatchString(projectSlug) {
return nil, fmt.Errorf("invalid project slug: %q", projectSlug)
}
baseURL := fmt.Sprintf("%s/api/0/projects/%s/%s/keys/", c.baseURL, orgSlug, projectSlug)
// ...
}
Apply the runtime regex to ALL slug-derived URL components (including orgSlug if it can ever be attacker-influenced), not just to config-time validation.
Credit
Reported by independent security research by Martin Brodeur.
The vulnerability can be exploited over the network without needing physical access. It is easy for an attacker to exploit this vulnerability. An attacker does not need any special privileges or access rights. No user interaction is needed for the attacker to exploit this vulnerability. The impact is confined to the system where the vulnerability exists. There is a low impact on the confidentiality of the information.
Exploitation attempts have been detected. Elevated vigilance and prompt remediation are advised.
Probability that this vulnerability will be exploited in the wild within the next 30 days.
We did not find any exploit available. Neither in GitHub repositories nor in the Exploit-Database.
- CVE-2026-47256Alias
Browse More
Continuously monitor your dependencies and get alerted when vulnerabilities like this one affect your stack.
Checkout DevGuard