Siri protocol cracked to run on any device, still requires iPhone 4S UDID
The Next Web reports that app developer Applidium, who you may remember from the VLC iOS app fracas earlier this year, has cracked the security protocols associated with Siri. Using Applidium's methods allows Siri to recognize voice commands from any device -- not just another iPhone, but also an iPad, or even potentially a Mac or Android device.
As of now the only thing keeping this from reaching a widespread audience is the fact that an iPhone 4S UDID is still required to get Siri to recognize inputs. Technically an enterprising developer with the proper know-how could tweak things to associate the iPhone 4S UDID with another device (as Applidium has done), but as The Next Web points out it would be very easy for Apple to track down such activity and shut it down.
Applidium has a post describing its methods as well as some of its other findings. Among the interesting things they learned: Siri sends compressed audio in Speex format (an audio codec associated with VoIP) of your every command to Apple's servers. Other data includes timestamps and confidence scores for each word spoken, all of which presumably helps out a great deal in deciphering the actual content of your commands. In other words, our assessment from a few weeks back appears to be correct: the device itself determines what you said, while Apple's servers take that data and extrapolate it into what you meant.
Applidium's work is just the latest bit of evidence that there shouldn't be any technical hurdles holding Siri back from more widespread deployment than just Apple's newest iPhone. Despite Apple engineers' comments that they "currently" have no plans to spread Siri beyond the iPhone 4S, it seems inevitable that the service will eventually expand to other devices in Apple's ecosystem.
Share
Categories
The Next Web reports that app developer Applidium, who you may remember from the VLC iOS app fracas earlier this year, has cracked the...
Add a Comment
What? You guys are totally wrong. The iPhone does not determine what you said, the servers which receive the audio determine what you said.
November 22 2011 at 1:22 PM Report abuse Permalink rate up rate down ReplyThis abbreviated description is a tad confusing. According to Applidium's own description, the "other data" you mentioned, confidence scores and timestamps, is sent back to the iPhone by the servers when using "text-to-speech" (surely they meant speech-to-text...) Since the transcription is done on the remote servers, not the phone, there is nothing on the phone from which to generate a confidence rating.
November 15 2011 at 12:44 PM Report abuse Permalink rate up rate down Reply"Siri sends compressed audio in Speex format (an audio codec associated with VoIP) of your every command to Apple's servers."
So... as a test of concept stated above, does this mean that if the phone is put into Airplane mode that Siri has lost her voice ??
I might be wrong, but Apple depreciated UDIDs as device identifiers. They use a different kind of identifier now.
November 14 2011 at 6:44 PM Report abuse Permalink rate up rate down ReplyThey deprecated them for developer use, but I'm not sure they have stopped using them internally.
November 14 2011 at 11:55 PM Report abuse Permalink +1 rate up rate down ReplyDeals of the Day
more deals- Acoustic Research Digital Photo Frame with iPod Dock for $50 + free shipping
- Apple iPhone 4 8GB for Verizon, AT&T, or Sprint for $50 + pickup at Best Buy
- Unlocked iPhone 4S 16GB for GSM (AT&T, T-Mobile) for $619 + free shipping
- Apple iMac Core i7 Quad 3.4GHz 27" w/ 24GB RAM, 2TB HDD for $2,677 + $29 s&h
- Used Apple Magic Mouse for $36 + $4 s&h
- Skullcandy Riot Earbud Headphones for $10 + free shipping
5 Comments