Secure online testing platform
Secure online testing, on whatever platform delivers the exam.
Aiseptor is the security platform underneath your online testing platform. It stops the AI cheating that runs outside the exam window, invisible overlays, on-device LLMs, remote access and proxy test takers, at the device and network layer. No webcam. No kernel driver. Live on any candidate device in about 30 seconds.
“Secure” used to mean a locked window and a camera.
Every online testing platform ships some security: a secure browser mode, a webcam feed, tab-switch detection, a plagiarism check. Against the cheating those were designed for, they work. They were designed for a candidate opening another tab or a friend in the room.
The cheating that doubled in a single year is different in kind. An invisible AI overlay draws answers on top of the screen and excludes itself from screen capture. An on-device LLM answers with no internet connection to inspect. A proxy test taker drives the session over a remote-access tool while the registered candidate sits on camera. None of it produces a browser event or a webcam anomaly, so the platform reports a clean session.
Four layers. Most platforms cover two.
The tools are usually compared as one list, which hides the point: they operate at different layers and defend against different threats. Match the layer to the threat, then stack.
| Layer | Covered by | What it sees, and what slips through |
|---|---|---|
| The person and the room | Webcam proctoring, identity checks, live proctors | Nothing about the device. A candidate reading an overlay looks identical to one reading the question. |
| The exam window | Lockdown browsers, the delivery platform's secure mode | Anything outside that one window. Overlays, local models and remote-access tools never touch it. |
| The device | Aiseptor | Detects the technique: screen-capture-exclusion flags, local inference, virtual displays, overlay and process signals. |
| The network | Aiseptor | Default-deny per session: AI endpoints, remote-access protocols and second-device pivots cannot connect. |
Aiseptor also covers the exam window itself with Aiseptor Secure Browser (macOS and Windows, in beta), so one platform spans the window, the device and the network. The person and the room stay with your proctoring or identity provider.
What it stops.
Four classes of tool now do their work entirely outside the exam window. Aiseptor does not chase their names; it targets what they all have to do.
Invisible AI overlays
Cluely, OpenCluely, Pluely, Nistaro and the next fork. Every one must set the OS screen-capture-exclusion flag to hide, and every one must reach a model to answer. Aiseptor flags the technique and blocks the path, so a renamed build is caught the same way.
How Cluely is detected →On-device LLMs
Ollama, LM Studio and llama.cpp answer with no internet connection to inspect. Local inference leaves a process, model-file and GPU-memory footprint on the device, which is where Aiseptor looks.
On-device LLM cheating →Remote access and proxy test takers
A paid expert drives the session over AnyDesk, RustDesk or a custom agent while the registered candidate sits on camera. Under a default-deny policy the remote-access protocol never connects, so the proxy never joins.
How proxy services work →Second-device pivots
A phone or second laptop on the same network relaying questions to a model. Aiseptor governs the exam device's network identity for the session, so the pivot has nothing to relay through.
Network-layer exam security →Works with the testing platform you already run.
Because enforcement happens at the OS and network layer rather than inside a plugin, Aiseptor does not need a per-platform integration to protect a session. It governs what the device can reach regardless of which application makes the request.
LMS-delivered exams
Canvas, Moodle, Blackboard or your own portal. The exam page opens in a normal browser, or in Aiseptor Secure Browser, while Aiseptor enforces the machine boundary underneath. Nothing about the exam changes.
Remote exam integrity →Assessment platforms and hiring
Coding screens and technical interviews on HackerRank, Codility, CodeSignal, CoderPad and similar. Platforms embed Aiseptor through a REST API and webhooks; hiring teams run it per interview.
For assessment platforms →Certification and licensure
At-home delivery without a test centre. Proxy rings and anti-detection VMs live at the device and network layer, which is the layer Aiseptor enforces.
For certification bodies →How to evaluate a secure online testing platform.
Six questions that separate the layer a vendor covers from the layer its marketing implies. Use them on us too.
Which layer does it enforce?
Window, person, device or network. Ask the vendor to say which, in one sentence. Most cover the window and the person.
What happens to a tool it has never seen?
A denylist of process names catches one named binary and goes blind at the next rename. Ask how an unknown overlay is caught.
Does it run on unmanaged, BYOD devices?
Kernel drivers and admin rights fail on a candidate's personal laptop. Ask for the install path and the removal path.
What does it collect from the candidate?
Webcam, microphone, keystrokes, screen recording, or none of them. Privacy law and candidate trust both turn on this.
Can it prove what it did?
A record of what was and wasn't reachable during the session is evidence. A suspicion score is an opinion.
Does it change the exam?
If it requires re-authoring content or moving platforms, the cost is the migration, not the licence.
Comparing specific products? The 2026 buyer's guide groups 15+ platforms by layer, and the alternatives pages go tool by tool.
Secure the session without surveilling the candidate.
Enforcement at the device and network layer only needs to know where the device is trying to go, not what the candidate is typing, saying or looking at. That removes the privacy exposure, bias risk and dispute surface of camera-based proctoring, and it is why Aiseptor can run in jurisdictions with strict candidate-privacy requirements.
- No webcam. No microphone.
- No keystroke logging. No clipboard capture.
- No screen recording. No personal files read.
- No footage captured; 90-day default record retention, configurable 7-365.
Tested on real sessions, not a demo.
Three signed pilots with assessment platforms, and two live runs of 100+ candidates each at Cornell University.
FAQ
Secure online testing platform questions.
What is a secure online testing platform?
A secure online testing platform is the combination of an exam delivery system (an LMS, an assessment platform, or a certification body's own portal) and the security controls that keep the session honest: identity checks, exam-window lockdown and, in 2026, enforcement at the device and network layer against AI tools. No single product covers all of it. Aiseptor is the device- and network-layer security platform; it works with whichever delivery platform runs the exam.
Is Aiseptor an exam delivery platform?
No. Aiseptor does not author, host or grade tests. It secures the candidate's device and network path while your existing platform delivers the exam: the exam page opens in a normal browser, or in Aiseptor Secure Browser, and Aiseptor enforces the machine boundary underneath. That is deliberate. Institutions keep the delivery platform their content, integrations and proctoring already live in, and add the layer it was never built to cover.
Which online testing platforms does Aiseptor work with?
Any browser-delivered assessment. Because enforcement happens at the OS and network layer rather than inside a browser plugin, Aiseptor does not need a per-platform integration to protect a session: it governs what the device can reach regardless of which application makes the request. Assessment platforms and LMS vendors can embed it through a REST API and webhooks; institutions can run it alongside their current platform without changing the exam itself.
What does Aiseptor stop that a lockdown browser or webcam proctoring does not?
The cheating that runs outside the exam window: invisible AI overlays such as Cluely and its forks (OpenCluely, Pluely, Nistaro), on-device LLMs like Ollama and LM Studio that need no internet, remote-access tools that let a proxy test taker drive the session, and second-device pivots. A lockdown browser sees one window. A webcam sees the person. Aiseptor detects the technique these tools share (a screen-capture-exclusion flag, a GPU and process footprint, a network path to a model) and blocks the path under a default-deny policy, so renamed and never-before-seen tools are caught the same way the known ones are.
Does it need a webcam or screen recording?
No. Aiseptor uses no webcam, microphone, keystroke logging or screen recording. Enforcement happens at the OS and network layer, so it only needs to know where the device is trying to go, not what the candidate is typing, saying or looking at. Because it captures no footage there is none to retain: post-exam records default to 90 days and are configurable from 7 to 365 days by the institution.
How is it priced?
Per exam session, usage-based and quoted against your volume. There are no seat licences and no annual contract is required to pilot. Every account starts with five free sessions.
Does Aiseptor replace proctoring?
It replaces a standalone lockdown browser (Aiseptor Secure Browser, in beta, locks the exam window on macOS and Windows) and it covers the device and network layer no proctoring service reaches. It does not verify identity, watch the room, or catch a phone or a paper note off-camera. Programs that need that keep a proctoring provider or an identity check for the person and run Aiseptor for the device.
Keep your testing platform. Secure the device.
Five free sessions, no credit card. Run one real exam and read the record of what the device tried to reach.