Nowadays I has been rapidly influencing all the industries as fast as possible suggest help space was sleeping mask just buses the people who have build and investing space have curious about AI what next and how they can use it for their purpose but we are now living in the beginning stage of a i most of us now thinking that I could do everything which human can do. Assumption is now we haven't achieved the peak of AI. There are some Pessimistic point in Ai which we are using today.
We can easily confused AI systems today. For example if you take any computer vision based face detection system to detect your face, open PC system it just put a glass in your face the AI system will confuse to predict fees this cold physical attacks on AI system.
What is adversarial attack?
Adversarial attacks are manipulate that aim to machine learning performance, course error in the model performance
You can make Bussiness queries by using the business queries section. You can find that at the bottom of this page
- You can make queries related to place ads, product placement, and advertising
- make project queries related to DIY project section (college and school students for their projects)
Adversarial attacks against computer vision
In the beginning of this article I have mentioned about physical adversarial attacks on visual data.
In 2018, group of researchers showed that by adding sticker on stop sign good able to fool computer vision system of a self driving car to mistake it for speed limit sign.
(picture from Google)
In real case adversarial attacks against for facial recognition system in protest where demonstrators use makeup and stickers to fool the surveillance cameras powered by machine learning algorithms.
- check here my previous tutorials
- ESP32-CAM Face detection|Face Recognition
- Machine to Machine Interactions
- mesh networking- explained
The adversarial attack against the text to speech
Text-based adversarial attacks involve making changes in the sequence of words the piece of text article, will cause miss classification errors in machine learning algorithms.
(Picture from Google)
Adversarial attacks against speech recognition systems
In a hypothetical adversarial attack, a malicious actor will carefully handle an audio file – say, a song posted on YouTube – to deem a hidden voice command. A human listener wouldn’t notice the change, but to a machine learning algorithm looking for patterns in sound waves it would be clearly audible and actionable. For example, audio adversarial attacks could be used to secretly send commands to smart speakers.
(picture from Google)
No comments:
Post a Comment