Intel introduced FakeCatcher, a solution to counter deep fakes with 96% accuracy

Recently, Intel has claimed that the company has finally come up with a solution to counter the growing influence of deepfake with its FakeCatcher. Deepfakes are self-made media where the original person in a video is replaced very precisely with another one.

The company is claiming that their FakeCatcher software is 96 percent accurate and performs its action by examining the flow of blood through pixels and giving results in less than a second.

The team behind this solution consists of Ilke Demir, a senior scientist at Intel Labs, and Umur Ciftci, from the State University of New York. The software is run with the help of the company’s software and hardware.

In comparison to other such detecting software, FakeCatcher works on the basis of PPG, a technique in which the light that goes into blood vessels or is deflected by them can be traced. giving it an edge over other detectors as it is based on real evidence.

As told by Ilke, the PPG technique has been known for some time, but applying it to detect a deepfake is being done by Intel.

The light signals are taken from thirty-two different points on the person’s face. Once the signals are contained, a map is designed based on the data received. And after the map has been designed, next up is the neural network to detect whether the given video is fake or not.

The counter to deepfake has come up at the best time possible. Lately, deepfake is being used by a number of vicious criminals in order to use the videos for immoral activities or blackmail people with them.

Forrester Research made its prediction two years ago, saying that the amount associated with such scams and threats was going to be over $250 million.

The threat came under the spotlight when it was used against the famous actor, Tom Cruise, and the well-known entrepreneur, Elon Musk. These personalities were used in a deep fake that made its appearance as advertisement. However, not all deep fakes are illegal. Many software houses, including Hour One, offer their deep-fake services to other businesses.

According to Intel, the company is still in its early stages with FakeCatcher and is still analyzing it thoroughly. Demir said that the company is not just settling with the PPG technique but also looking at other areas, such as the gaze method to detect if the video is genuine or synthetic.

However, detection is not an easy thing, as it has been observed that the results are affected in some cases where someone’s gender or skin tone makes it difficult, leading to a difference of over ten percent.

Demir said that currently the data used is based on thirty to forty people and still has a long way to go. Whereas, Rowan Curran from Forrester Research said that they are part of an evolutionary race to detect whether the media under observation is real or produced by someone else. He further added that Intel’s solution can play an important role in the future if the results are not affected by any of the factors.

Read next: Despite the global economic situation, Amazon and other cloud service providers did well in Q3 of 2022
Previous Post Next Post