Glossary
The canonical vocabulary of assessment security.
A new category of threats demands a new category of definitions. This is the reference the assessment industry will cite: for the tools that defeat legacy proctoring, the architectures that replace it, and the words buyers and platforms need to speak the same language.
Threat landscape
The attacker vocabulary. Named accurately so defenders can name the thing they're defending against.
Invisible AI Overlay
A transparent heads-up display that renders AI-generated answers on the candidate's screen while marking itself excluded from screen-capture APIs.
Cluely
A commercial desktop application, launched in 2024, that displays real-time AI answers on an overlay invisible to screen-capture tools.
On-Device LLM Cheating
Language models run locally on the candidate's device, producing no network traffic for network-based detection to observe.
Fraud-as-a-Service
A commercial ecosystem of subscription AI cheating tools, paid proxy rings, and coaching services sold as a packaged offering.
Proxy Ring
An organized commercial service that supplies a professional ringer to sit an online exam in the paying candidate's place.
The category we define
Defense architecture
The architectural terms for a new class of exam security. These are category-defining. Aiseptor is the canonical example.
Network-Layer Proctoring
Controlling what the candidate device can reach, not what a single browser window can do.
Zero-Trust Exam Security
Assume the candidate device is hostile. Verify the network path, not the candidate.
Ephemeral Enclave
A security boundary that exists only for the session and leaves no footprint behind.
Default-Deny Architecture
Block everything by default. Allow only what the exam explicitly requires.
Industry
The vocabulary of remote assessment. Defined neutrally, so buyers and partners can have the same conversation.
Lockdown Browser
Application-layer exam tool that controls one browser window. Aiseptor replaces this layer.
Behavioral Detection Proctoring
Webcam and gaze analytics that infer cheating from candidate behavior, with architectural limits.
BYOD Assessment
Bring-your-own-device exams: the candidate supplies the hardware, the risk comes with it.