IBM Tests Audio-Based Large Language Model to Hijack Live Conversations
With the idea in mind to “audio-jack” a live call-based banking transaction, security researchers were successful in inserting cybercriminal-controlled account details. Deepfake audio is nothing new… but it is getting very advanced. So much so, that security researchers at IBM Threat Intelligence were able to test out a hypothesis as to whether it’s possible to perform an audio-based “Man in the Middle” attack.