This new thing is a trusted third party between the media You must and advertisers.
—Trusted Third Party—
Let’s go back to the previous example. A is the advertiser? B is the media? and the trusted third party is C. C’s responsibilities are very important.
The advertiser (A) needs to encrypt his own 5?000 mobile phone numbers and send them to a third party (C)? and then C will help mix in 45?000 obfuscated mobile phone numbers.
Then C helps A to match You must
Data with the media (B). After the matching is these systems can automate actions completed? B sends its own data results (individual-level user attribute data) to C in encrypted form.
C removes (strips) the encrypted attributes of the obfuscated ID from the encrypted attributes sent back by B? and then calculates the encrypted attributes of those people who can be matched among the advertiser’s 5?000 mobile phone numbers? and summarizes the common characteristics of these people. These common characteristics no longer carry personal attributes? but are statistical data? so they no longer involve the ID of advertiser A’s consumers. And these openai said the models can even common characteristics are calculated from the encrypted attributes sent by media B? so they are also encrypted.
The third party
C then returns these common features to the media B. The media decrypts these encrypted common features and then looks for people with the same or similar common features to help advertisers place advertisements.
Figure 3: Anonymous secure transaction with a trusted and aero leads data third party (see attachment)
— Homomorphic Encryption —
Why C calculates the encrypted individual attributes sent by media B. Can it still be calculated after encryption? The answer is yes. Because privacy computing has another technology called “homomorphic encryption”. The so-called homomorphic encryption means that the result of calculating the original text (plaintext) is exactly the same as the result of calculating the ciphertext after the plaintext is encrypted.
