Assessment that survives being unsupervised

Design for it rather than trying to prevent it.

Unsupervised online assessment cannot reliably establish who completed it or what they consulted, and elaborate proctoring arrangements carry their own costs in privacy, access and false accusations. The more productive route is designing assessment where those questions matter less.

Tasks drawing on the candidate's own context, workplace or specific experience are difficult to complete on somebody else's behalf and difficult to copy. So are tasks requiring a decision to be justified rather than a fact to be produced, and tasks that build on something the candidate submitted earlier.

Where a high-stakes judgement genuinely requires supervision, that should be concentrated in one properly supervised component rather than spread across everything, which is both cheaper and more defensible than partial invigilation of many small tasks.

The other option is a conversation. A short structured discussion about submitted work establishes authorship more reliably than any technical measure, and it takes minutes per candidate.

The other advantage of assessment drawn from the learner's own context is that it is more useful to them. A task requiring somebody to apply the material to their actual workplace produces something they can use, which changes their engagement with it, and it is also the form employers value when they are funding the training.