Ability to access private data from MTurk HIT Interface

Question

When using MTurk to annotate data, is there any way to host that data so that it is private and only accessible via the HIT Interface built for the specific task? In other words, using S3 or a different service, is it possible to leverage some aspect of the MTurk framework to validate that a request for data is coming from an MTurk worker using our HIT Interface? In this situation, I can not make the data public. It'd be ideal if there were some way for AWS / MTurk to get access to a private S3 Bucket in such a way that it can 'check out' resources from my bucket (i.e. a video) when sending a task to a worker, so that there's no requests for that data in the HIT interface's code.

Answer

Hello,

Making use of S3 [pre-signed URLs](https://docs.aws.amazon.com/AmazonS3/latest/userguide/ShareObjectPreSignedURL.html ) to generate URLs for the private objects in your S3 bucket can prove useful to you. Pre-signed URLs are temporary URLs that provide time-limited access (in which you can [customize](https://docs.aws.amazon.com/AmazonS3/latest/userguide/using-presigned-url.html#PresignedUrl-Expiration:~:text=the%20download%20fails.-,Limiting%20presigned%20URL%20capabilities,-The%20capabilities%20of) to your liking) to S3 objects. These URLs can be used by the Mechanical Turk HIT interface to directly access the private resources. After the URL's have been generated, you can then embed them into your HIT's. When a worker accepts the HIT, a pre - signed URL will appear in the HIT instructions. Only workers that accept the HIT will be granted access.

In terms of the [validation process](https://docs.aws.amazon.com/AWSMechTurk/latest/AWSMechanicalTurkRequester/RetrievingResults.html), you can add code in your HIT interface to validate that the requests for data are coming from workers on MTurk. This could involve checking the MTurk worker ID or using other methods of worker verification.

Ability to access private data from MTurk HIT Interface

相关内容