I assume it won't work for Skype then?
Am I wrong, or will the (partial, at least) peer-to-peer nature of Skype mean this won't work with it?
Security researchers in the States say they have developed a cunning new method of "fingerprinting" voice calls that could offer a route to trustworthy caller ID and a barrier against so-called "vishing" or voice phishing. The tool is called PinDr0p, and works by analysing the various characteristic noise artifacts left in …
"PinDr0p needs no additional detection infrastructure; all it uses is the sound you hear on the phone."
Or rather, the noise you don't hear..
"“They’re not able to add the kind of noise we’re looking for to make them sound like somebody else,” says Patrick Traynor, GIT compsci prof. “There’s no way for a caller to reduce packet loss. There’s no way for them to say to the cellular network, ‘Make my sound quality better.’”"
Presumably it's possible to trick it into believing a reduction in quality is there though - for example deliberately introducing tiny gaps into a non-VOIP call to trick the system into thinking it's detecting packet loss? Not sure to what end you'd want to do that, but...
This would only work if the perps are able to get at the packet stream and selectively remove the packets themselves. It won't work if they just add bits of "silence"; that does not result in a packet drop.
Not impossible, but you'd probably need to route the call via your own VoIP network (since private ISDN networks are few and far between) to achieve anything.
or does it sound like this has the potential to flag a call as suspicious, even false, due to the routing the mobile network uses, which is completely out of your control.
while conmen may do this on purpose, there is no reason why a network couldn't route a call anywhere if it turned out to be cheaper for them.
I worked with voice-print technology many years ago when we barely had computers. It can be very accurate to analyze the noise under the audio for specific characteristics. I don't think artificially induced moments of silence would fool the algorithm.
I'm sure it's a lot easier today.