Elastic APM Agents

Skip to content What this pays sorona.xyz pays 1.25× the program's reward, up to +$1,000 per finding, plus a complexity bonus of up to $1,000 for especially complex or well-chained bugs. Finding │ Program pays │ Through sorona.xyz Critical (Observability - APM Agents row) │ $7,000 │ $8,000 High │ $3,000 │ $3,750 Medium │ $1,500 │ $1,875 Low │ $700 │ $875 Submit a vulnerability CVE history 0 Last 12 months 6 Lifetime Info disclosure 40 % Code execution 20 % Denial of service 20 % XSS 15 % Latest advisories Loading the feed Overview APM agents are among the highest-privilege pieces of third-party code in a modern application estate, and they ship in eight languages covering essentially the whole enterprise stack: Java, .NET, Go, Node.js, Python, Ruby, PHP and browser RUM. They are pulled from Maven Central, npm, PyPI, NuGet, RubyGems and Packagist, so adoption is measured in package-manager downloads rather than instance counts, and they are installed by default along Elastic Observability's onboarding path. The security significance is far larger than the star counts suggest: an APM agent is loaded into the same process as the application, sees every HTTP request, every database query and often every header and body before redaction, and holds a credential to the observability backend. A sanitisation bypass silently exfiltrates production secrets into a logging cluster that usually has a much wider read audience than the application itself, and an instrumentation or memory-safety flaw is code execution inside the instrumented service. Elastic N.V. pays the family as one asset on HackerOne against a bound row running from $700 for a Low to $7,000 for a Critical, awarded at triage. sorona.xyz adds 1.25x on top of whatever Elastic awards, up to $1,000 more, plus a complexity bonus of up to $1,000 for a well-chained finding. Response times ~11 hours First response ~6 days Triage ~43 days Bounty In scope The declared asset is Observability - APM Agents, bounty-eligible up to critical severity, and it covers the family rather than one flagship: apm-agent-java , -dotnet , -go , -nodejs , -python , -ruby , -php and -rum-js No proprietary carve-out applies here. Unlike Elasticsearch and Kibana there is no x-pack/ directory in any agent repository, so the whole covered artefact is open source under Apache-2.0, BSD or MIT, and Elastic's invitation reaches all of it: 'If we wrote code with a bug in it, we want to know about it! Our code is open so use that to your advantage!' Each agent's instrumentation surface in full: the frameworks and libraries it hooks, HTTP servers and clients, database drivers, message queues, the transaction and span model, and distributed-tracing context propagation The attach mechanisms that get the agent into the process: JVM premain and runtime attach, the .NET profiler, Node loader hooks, monkey-patching in Python and Node, and the PHP extension Configuration handling: environment and file config parsing, the central-configuration channel agents poll from APM Server and apply at runtime, and the sanitisation rules that decide which fields, headers and bodies are captured The transport to APM Server: payload serialisation, compression, TLS and certificate verification, and how each agent stores the secret token or API key it authenticates with apm-agent-rum-js , which executes in the browsers of the customer's own end users on pages the customer serves Out of scope Reports that reach Elastic by any route other than its own HackerOne program, which the policy makes a condition of bounty eligibility Third-party systems, domains, services and components, which fall outside the policy even when an agent depends on them Previously known vulnerable dependencies reported without a working proof of concept Issues that only affect EOL agent versions rather than a currently maintained release Volumetric denial of service, which the policy rules out permanently rather than case by case Findings requiring administrative or root privileges on the host before they can be reached Server-side issues in APM Server itself, which is a separate scope asset with its own bounty row Most-valued findings A redaction or sanitisation bypass that ships production secrets, authorization headers, request bodies or database parameters into the observability cluster, the dominant class at 40% and the one that turns a logging backend into a credential store Code execution inside the instrumented application through the attach or instrumentation mechanism, whether that is bytecode weaving, profiler attach, or a loader hook Abuse of the central-configuration channel to make an agent apply attacker-influenced settings at runtime inside a production process Credential or secret-token exposure from the agent's own storage or from its transport to APM Server, including TLS verification weaknesses Cross-site scripting reachable through apm-agent-rum-js , which executes in end users' browsers on the customer's own origin An instrumentation path that reliably crashes or deadlocks the host application, provided the effect does not depend on traffic volume Program rules Submit through sorona.xyz first. We validate, coordinate with the program's security team, and add the payout uplift on top of whatever it awards. Exclusive routing through hackerone.com/elastic is a condition of bounty eligibility Reproduce on a currently maintained agent release; EOL agent versions are out of scope Findings that require root or administrative OS privileges before they work are excluded Volumetric denial of service is never eligible for a bounty reward A known-vulnerable-dependency report needs a working proof of concept showing the path through the agent The reward amount is at Elastic's sole discretion, and the published row is a minimum rather than a fixed price Program stats 891 Reports resolved $986k Bounties paid 79% Response efficiency Disclosure Coordinated by Elastic, which self-assigns CVEs as a CNA and publishes them as Elastic Security Advisories alongside the NVD record. Bounty eligibility is conditional on exclusive submission through Elastic's own program, so file there first; Elastic awards bounty-eligible reports at triage rather than waiting for the fix to ship. Safe harbor Sorona will not pursue legal action against you for good-faith research submitted to us in line with this policy. We cannot speak for the upstream project, its vendor, or any third party, and we do not grant authorization on their behalf. You are responsible for understanding and following the laws that apply to you, wherever you are, and for staying within any scope the project has defined. Submit a vulnerability