Ultrasonic to a phone or Alexa/Siri/etc, connect to an unsecured network, send data to a neighbor’s smart TV which is connected to Internet, Bluetooth or other to a phone
DNS will tell you the server name and address, which would just be some server owned by the company. Nothing weird there unless they have the chutzpah to name it something telling. They could even bypass DNS entirely with hardcoded IP addresses.
Timing wouldn’t be a great indicator either if they aggregate requests.
They could slide anything nefarious in with daily software update checks or whatever other phone-homing they normally do, and without deep packet inspection or reverse engineering the software, it would be very difficult to tell.
I don’t think Wireshark can do deep packet inspection, can it? Assuming the client is using SSL and verifying certs, maybe even using cert pinning?
Size would be a big indicator if they’re sending full voice recordings, but not if they’re doing voice recognition locally and only sending transcripts, metadata, or keywords.
I’ve never actually done this kind of work in earnest, and my experience with Wireshark is at least a decade out of date. I’m just approaching this from the perspective of “if I were a corporate shitbag, how would I implement my shitbaggery?”
Just spitballing here but you might be able to try and correlate the amount of data sent with how much real life activity there was. Say, have silence for a week around the TV then play recorded speech near it for a week and see if that changes the frequency or size of the data being sent back home. Then do this for random 1/2/3 day periods. If offline text to speech is as crap as I’ve heard then the increased data transfer should stick out pretty clearly.
First, someone would be able to prove that communication is happening. Second, if the keys are stored locally, and the original packets saved, the encryption can be reverse engineered.
Encryption prevents man in the middle attacks. If you have one of the ends, you can usually get the data. If you have the device that’s doing the encryption of the data, and you have the encrypted data, you can decode the data. It’s just a matter of getting through obfuscation at that point.
The reason this hasn’t been done yet is that it’s not happening yet. CMG was lying in their advertising.
I’m sure it will show HTTPS traffic outbound from your TV.
I’m sure it will show no traffic whatsoever if you don’t connect your TV to your network
There’s a dozen ways they could jump the air gap.
Ultrasonic to a phone or Alexa/Siri/etc, connect to an unsecured network, send data to a neighbor’s smart TV which is connected to Internet, Bluetooth or other to a phone
But this would be proven then?
And with DNS requests and timing you should be able to figure whats in those packets.
Sorry if this is a noob question, but…how?
DNS will tell you the server name and address, which would just be some server owned by the company. Nothing weird there unless they have the chutzpah to name it something telling. They could even bypass DNS entirely with hardcoded IP addresses.
Timing wouldn’t be a great indicator either if they aggregate requests.
They could slide anything nefarious in with daily software update checks or whatever other phone-homing they normally do, and without deep packet inspection or reverse engineering the software, it would be very difficult to tell.
I don’t think Wireshark can do deep packet inspection, can it? Assuming the client is using SSL and verifying certs, maybe even using cert pinning?
Size would be a big indicator if they’re sending full voice recordings, but not if they’re doing voice recognition locally and only sending transcripts, metadata, or keywords.
I’ve never actually done this kind of work in earnest, and my experience with Wireshark is at least a decade out of date. I’m just approaching this from the perspective of “if I were a corporate shitbag, how would I implement my shitbaggery?”
The answer is: it wouldn’t. You’re right on the money, you couldn’t do anything other than speculation.
Just spitballing here but you might be able to try and correlate the amount of data sent with how much real life activity there was. Say, have silence for a week around the TV then play recorded speech near it for a week and see if that changes the frequency or size of the data being sent back home. Then do this for random 1/2/3 day periods. If offline text to speech is as crap as I’ve heard then the increased data transfer should stick out pretty clearly.
First, someone would be able to prove that communication is happening. Second, if the keys are stored locally, and the original packets saved, the encryption can be reverse engineered.
Encryption prevents man in the middle attacks. If you have one of the ends, you can usually get the data. If you have the device that’s doing the encryption of the data, and you have the encrypted data, you can decode the data. It’s just a matter of getting through obfuscation at that point.
The reason this hasn’t been done yet is that it’s not happening yet. CMG was lying in their advertising.
That’s not how that works lol