The language of autonomouspenetration testing.

Glossary entries

Autonomous pen-tester

A platform that plans, executes, and reports a penetration test end-to-end without a human driving each step.

An autonomous penetration tester is a platform that accepts a natural-language engagement goal and scope, synthesises a plan, executes the plan against authorised targets, captures live evidence, and produces a report. Unlike a scanner it chains exploits and adapts on findings. Unlike a human consultant it runs continuously and deterministically.

Agentic AI

AI systems that pursue goals by taking real actions — calling tools, running commands, reading results, deciding what to do next.

Agentic AI describes systems where a language model is embedded in a loop that can invoke tools, observe real-world state, and adjust behaviour on each observation. In offensive security this means running scanners, parsing output, launching exploits, capturing sessions, and feeding evidence back into the next decision — all inside an approval and sandbox perimeter.

Sandbox-verified exploit

An exploit that has actually been detonated inside an isolated environment and produced a captured artifact.

A sandbox-verified exploit is a finding backed by a real execution trace. The platform ran the exploit against the target inside an isolated container. The payload fired. An artifact (shell session, dumped credential, captured response) was recorded. No sandbox artifact means no finding. This is the line between an action-worthy security report and a language model's guess.

Signed evidence chain

A cryptographically sealed record of every action, artifact, and session log from an engagement.

A signed evidence chain hashes each event and artifact from an engagement into a tamper-evident structure (typically a Merkle tree) then signs the root with the operator's key. Auditors can verify the entire chain with a single signature check. A pentest report shipped with a signed evidence chain is a cryptographic artifact, not a PDF of claims.

Human-in-the-loop (HITL) approval

A mandatory operator sign-off required before any destructive or privilege-changing step.

Human-in-the-loop approval is a gate in the engagement plan that blocks the platform from executing a specific action (lateral movement, privilege escalation, data exfiltration, destructive exploit) until a human operator explicitly approves. It keeps autonomous agents inside the rules of engagement and inside the policy your insurer and your customers expect.

Zero-day discovery

Finding a previously unknown vulnerability through fuzzing, crash triage, and proof-of-concept synthesis.

Zero-day discovery goes beyond matching known CVEs. The platform generates semi-random inputs against a target, captures crashes, triages each crash for exploitability, and synthesises a proof-of-concept where one exists. It surfaces vulnerabilities that no public database has catalogued yet.

Continuous security validation

Running the same engagement plan on a schedule so every change to your environment is re-tested.

Continuous security validation turns a one-off pentest into a running check. The same plan re-executes on a schedule or trigger. Regressions show up within hours instead of between annual engagements. Evidence from every run is retained so trend analysis is possible.

N-day vs zero-day

N-day vulnerabilities have a published CVE; zero-day vulnerabilities do not yet.

An N-day vulnerability is a known, catalogued weakness with a CVE identifier. The fix (and usually the exploit) is public. A zero-day vulnerability is one the vendor has not yet patched and that no public database has indexed. DXSense discovers both: N-day via matching, zero-day via the fuzzing and PoC-synthesis pipeline.

SOC 2 vs ISO 27001

SOC 2 is a US audit of controls against service-org principles. ISO 27001 is an international certification against an ISMS standard.

SOC 2 (by AICPA) and ISO 27001 (by ISO/IEC) are the two dominant enterprise security attestations. SOC 2 Type II audits a service organisation's controls across Security, Availability, Processing Integrity, Confidentiality, and Privacy over a period. ISO 27001 certifies an Information Security Management System against the ISO/IEC standard. Enterprise buyers frequently require one or both in their vendor security review.