Can AI prevent “them” from eavesdropping on you?

You may have even encountered such a situation:

When you open an app on your phone, the first thing you see is what you talked about with your friends or family just a few minutes ago. The first time, you will feel surprised. But after a long time, you will get used to it and it will no longer be strange.

One of the main culprits is the microphones that are now everywhere and "eavesdropping" on you all the time. They are embedded in mobile phones, TVs, watches and other devices, transmitting your voice to neural networks and artificial intelligence systems in real time, helping the recommendation system to make "customized" push services for you.

(Source: Pixabay)

So, how can we avoid this? There are many ways on the Internet, such as playing a completely unrelated song, turning up the volume to the maximum, and actively creating some noise; for example, not authorizing long-term microphone access, but choosing to take a one-time authorization approach.

Now, a research team from Columbia University in the United States has proposed a new method: they have developed an artificial intelligence system that can avoid the occurrence of "eavesdropping" incidents by simply playing an extremely slight sound produced by the system in the room.

In other words, this system will disguise the sound of people's conversations so that they cannot be heard by microphones or other monitoring systems without affecting normal conversations.

The related research paper, titled “Real-Time Neural Voice Camouflage”, has been published on the preprint website arXiv.

Moreover, the researchers also stated that this artificial intelligence system can be easily deployed on hardware such as computers and mobile phones, and can protect your privacy at all times.

Beating AI with AI

Problems that are exploited by artificial intelligence should be solved by artificial intelligence.

While the team’s work on jamming systems like microphones is theoretically possible in the field of artificial intelligence, getting it fast enough for practical applications remains a difficult challenge.

The problem is that jamming a microphone listening to people talking at one particular moment may not disrupt the conversation a few seconds later. When people talk, their voices change constantly as they speak different words and at different speeds, making it nearly impossible for machines to keep up with the fast pace of human speech.

In this study, the AI ​​algorithm was able to predict the characteristics of what a person was going to say next and had enough time to produce an appropriate whisper.

(Source: Pixabay)

Carl Vondrick, an assistant professor of computer science at Columbia University, said the algorithm can block rogue microphones from listening to conversations with 80% efficiency, and it works even when people are unaware of information such as the location of the rogue microphone.

To do this, the researchers needed to design an algorithm that could break neural networks in real time, that could be generated continuously while speaking, and that could work for a large portion of a language’s vocabulary. No previous work had met all three requirements at the same time.

The new algorithm uses a type of signal called "predictive attacks" that can interfere with any words that the automatic speech recognition model is trained to transcribe. In addition, when the attack sounds are played over the air, they need to be loud enough to interfere with any rogue "eavesdropping" microphones that may be far away. The attack sounds need to travel the same distance as sound.

The AI ​​system achieves real-time performance by predicting future attacks on signals or words, conditioned on two seconds of input speech.

At the same time, the research team also optimized the attack so that its volume is similar to normal background noise, allowing people to talk naturally in the room without being monitored by systems such as microphones.

Furthermore, they successfully demonstrated the effectiveness of this approach in real rooms with natural ambient noise and complex shapes.

But Vondrick also said that currently the system is only effective for most English words, and they are applying the algorithm to more languages.

What can we do for ethical AI?

Some people say that in the era of big data, our personal information is in a "naked" state, and there are more and more smart devices around us that are peeping into our personal privacy.

If this research result can be applied to more languages ​​and implemented in more scenarios in the future, it may help us avoid being exploited by various forms of artificial intelligence.

As Jianbo Shi, a professor at the Department of Computer and Information Science at the University of Pennsylvania, commented: They raise a new question: how can we use artificial intelligence to protect us from being exploited by unconscious artificial intelligence?

At the same time, Shi also suggested that in future work, researchers need to "consciously" think about the impact of artificial intelligence on humans and society from the initial design stage, and no longer ask what ethical AI can do for us, but what can we do for ethical AI?

Once we believe in this direction, research on ethical AI will be full of interest and creativity.

Original link:

https://arxiv.org/abs/2112.07076

https://techxplore.com/news/2022-04-spying-ai-block-rogue-microphones.html

https://c3.ai/glossary/artificial-intelligence/ethical-ai/

Academic headlines