In recent years, the use of deepfake technology to create fake videos and images has gained widespread attention, for example, with issues such as revenge pornography or the proliferation of GAN-generated social media profile images.
The Emerging Threat of Synthetic Audio-Based Fraud
However, another emerging threat is the use of synthetic audio. Although this can potentially be put to creative and positive uses, it can and is also used to commit fraud.
Synthetic audio, which is created using artificial intelligence and machine learning algorithms, has the potential to deceive individuals and organizations, leading to significant financial losses and reputational damage. In this post, we will discuss recent examples of synthetic audio-based fraud attacks and explore the technological measures that can prevent them.
One recent example of synthetic audio-based fraud is the use of voice cloning technology to impersonate executives and other officials. This was demonstrated in a well-publicized incident where criminals used synthetic audio to impersonate the CEO of a UK-based energy firm, convincing the company’s accountant to transfer $243,000 to their account. The criminals created a synthetic voice that closely resembled the CEO’s and used it to instruct the accountant to make the transfer. The fraud was discovered only after the real CEO called to inquire about the transfer.
Another example of synthetic audio-based fraud is the use of voice phishing scams. In these attacks, criminals use synthetic audio to impersonate bank officials, IRS or HMRC agents, and other authority figures to trick individuals into revealing sensitive information or making unauthorized transactions. According to the FBI, voice phishing scams cost US businesses over $2 billion last year.
Preventing Synthetic Audio-Based Fraud
So, how can organizations protect themselves from synthetic audio-based fraud?
One effective measure is the use of advanced voice biometric authentication. Voice biometric authentication uses the unique characteristics of an individual’s voice to verify their identity. By comparing the voiceprint of the speaker against a stored voiceprint, the system can determine if the speaker is who they claim to be. This technology has been used successfully by ValidSoft’s customers, including global banks and other enterprises, to detect and prevent fraud.
Another effective measure is the use of a standalone synthetic audio detection capability. This technology uses some of the same principles as voice biometric authentication to determine whether a stream of audio/speech is coming from a real human or has been synthetically generated. ValidSoft leads the way in this developing area and has brought to market technology that can detect synthetic audio reliably, which provides a powerful layer in the ongoing arms race against AI-based fraud techniques.
As synthetic audio technology continues to advance, it is likely that we will see more sophisticated fraud attacks that use this technology. However, organizations can reduce the risk of falling victim to these attacks by taking proactive measures to protect themselves and by deploying detection capabilities within their contact centers and other customer engagement channels.
To discuss ValidSoft’s synthetic audio detection capabilities, please get in touch!
Written by Alexander Korff, COO.
