Genius set a trap to prove Google stole song lyrics

Shawn Knight

TechSpot Staff
Staff member

Google in 2014 added song lyrics to its search results, meaning those searching for the words to a particular song often times don’t have to visit a third-party site to find what they are looking for. That’s been a bit of an issue for third-party lyric sites like Genius, and in more ways than one.

With fewer people navigating to lyric sites, traffic has predictably suffered. But Genius Media Group Inc.’s bigger complaint is that Google is allegedly lifting its lyrics and posting them without any credit.

Genius tells The Wall Street Journal that it provided Google with evidence of the theft in 2017 and again earlier this year. But how does one prove theft in this area? By setting up a trap.

Genius infused its lyrics with a clever watermarking system that used a pattern of straight and curly apostrophes that, in Morse code, spelled out the words “Red Handed.” In time, Genius said it found more than 100 examples of songs on Google that featured this exact same formatting.

The search giant claims it sources lyrics found in its “information panels” from partners and thus, doesn’t generate them on its own. Once The Journal’s article went live, Google issued a second statement saying it was investigating the matter and would cut ties with partners that were “not upholding good practices.”

Lead image credit: song writer holding pencil by panitanphoto

Permalink to story.

 

Evernessince

TS Evangelist
It's also possible to create a system that automatically tags clients by making some small inaudible variations like a change in volume, pitch, ect. The system would make a copy for each client, adjust the song based on a key, and compare those changes to the original through the use of an algorithm which can deduce client information based on the key. Essentially, you'd be encoding the client information specifically to that client's copy.
 

GeforcerFX

TS Evangelist
It's also possible to create a system that automatically tags clients by making some small inaudible variations like a change in volume, pitch, ect. The system would make a copy for each client, adjust the song based on a key, and compare those changes to the original through the use of an algorithm which can deduce client information based on the key. Essentially, you'd be encoding the client information specifically to that client's copy.
your talking for the actual music itself right? no idea how you would change the pitch or volume in text.