Words we use

The terms you meet in the console and in these docs, one line each.

The console uses a few words of its own. Each is explained here in one or two lines, with a link to the page that uses it.

​

Vishing simulation

Word	What it means	Where it appears
Attacker	The caller persona: a name, a role, a company, an opening line and a voice. One attacker can be used by many campaigns.	Create an attacker
Campaign	One run of calls: an attacker, a scenario, the people to call, a schedule and an authorisation record.	Build a campaign
Scenario	The story the attacker tells, for example an IT help desk reset or a bank fraud team. Seven single-call scenarios are built in.	The scenario library
Sequence	A chain of calls to the same person over days, where each call builds on the last. The phases are recon, elicitation, rapport, exploitation and exit.	Six ways to run a campaign
Test call	A call you answer yourself, in your browser. Nobody on your staff is dialled.	Take a test call in your browser
Rehearsal	The attacker runs its script against a simulated person several times, so you see how often it gets what it wants.	Six ways to run a campaign
Kill switch	A workspace-wide stop. When it is on, no campaign in the workspace places a call.	Pause, resume, stop everything
Warm-up	A short text or email sent to the person a few minutes before the call, so the call refers to something they have already seen.	Every message DeepBlocker sends to your staff
Coaching	At the end of a failed call, a separate DeepBlocker voice tells the person it was an exercise and what to watch for next time. Off by default.	Coaching at the end of a failed call
Verdict	The result per person: Passed, Failed, Reported, or Not reached.	Read your results
Risk score	A number from 0 to 100 per call, shown as “Risk 72/100”. Higher means the person gave more away.	Read your results
Detector caught	Whether DeepBlocker’s own deepfake detector flagged the cloned voice on that call. “Not scored” means the detector has not finished.	Read your results
Fair testing	The workspace rules that protect staff: cooling-off days, calls per quarter, quiet hours, new joiners, excluded people.	Fair testing rules
Minute	One minute of a simulated call, counted against your yearly licence.	How minutes work
Workspace	Your organisation’s space in the console. It holds your attackers, campaigns, people, voices, reports and licence.	Your workspace

​

Agent Assurance

Word	What it means
Rules of engagement	The signed record for one agent: who authorised the test, the dates, the scope, the test windows, the allowed and excluded techniques, and who to call if something goes wrong.
Sandbox	A safe copy of your agent’s tools. In the DeepBlocker sandbox, the agent’s actions hit our fake tools, so we record every action and nothing real happens.
Test conversation	One conversation in which our attacker tries to make your agent break a rule, over one channel. It stops after at most 30 messages.
Finding	One rule that was tested, with the level reached and the evidence.
Level	How far the agent went on one finding: 0 not prohibited, 2 blocked by a control, 3 attempted, 4 executed past a control that was not enforced, 5 executed with proof.
Exposure band	The overall result of a report: NONE, LOW, MODERATE, HIGH or CRITICAL. A test with no sandbox is capped at MODERATE, because we only saw words.