It’s remarkable to me how many competent programmers with years or decades experience in this industry don’t understand —- If you’re using AWS, Amazon has access to ALL of the data you put on AWS.
Not that they 'can' or 'want to', given the current state of technology they absolutely have to have access to all your data for AWS to function.
There isn’t currently a feasible technical way to work around this. And to head off all the ‘but FHE’ comments, see the ‘currently feasible’ above.
Access records for public services have a very detailed iam audit trail that logs people who accessed what at what time, and service teams don't get to just jump around that. Maybe they can see some metadata but certainly not actual data in an S3 bucket somewhere.
Even with ‘enclaves’, from what admittedly little I know about them, you still have to have the key to decrypt things on the machine somewhere, which means whoever is running that machine for you has access to your unencrypted data, and we’re back where we started.
The comments above indicating 'well someone has access' - yea, obviously, it's data hosting. Someone has access.
But the amount of conspiracy here is frustrating.
Amazon will play very aggressively within the bounds of the law, meaning, if they can glean public info about something, or look at their own sales data for a product, they will do that.
But to look at s3 data would risk the entire empire.
It's rational for people to be a bit skeptical, and so Walmart can say 'no data on AWS' but it's also an easy thing to do.
Now - is it possible that new retail PM, who used to be an AWS PM, and who for some reason still had access to things he shouldn't - went ahead and did that? That could happen. And maybe his boss finds out and looks the other way but calls IT and tries to have the loophole closed quietly. Etc.
As a policy are they trying to copy your product and even ask you for information and aggressively pursue customer data? Yes.
As a policy are they looking at your S3/ec2 data - no.
But individual actors are individual actors, in a company of 100 000 people, some will go astray.
They are pushing their 'white label' stuff agressively, I have no doubt the PM's have zero qualms about using Amazon.com sales data to their advantage.
But I also submit that retail PM's actually getting access to private S3/EC2 is totally rubbish, at least by any policy or scale.
They could be sued for billions in each case of that breach, and the resulting PR fallout would be impossible.
Imagine you are the VP of AWS - you make all the profit for Amazon.
Are you going to somehow allow some dirty Retail PM access to your customers data?
When your customer finds out, and tells the world, and it gets in the press, what happens?
If your ABC startup had evidence that Amazon was creeping on your data as policy, you'd have to dump them instantly.
They could say goodbye to every government contract.
If you are Bezos - would you risk the entire Brand and the cash-cow to move some low-margin pair of shoes and USB hub?
So no, I think the firewall between AWS and Retail is systematically legit.
I was certainly naive when I heard about other big retailers who would refuse to allow any subcontractors to use AWS. "Surely Amazon has a Chinese wall" to prevent that kind of data sharing, I thought. Never underestimate the lack of morals in business is the right answer I guess.