The paper presents a cryptographic method to protect privacy in social networks by reducing the correlation between publicly shared content (e.g., images) and sensitive hidden information (e.g., names of people in the images).
The key ideas are:
Randomization: The hidden content (e.g., name) is transformed into a randomized message 'u' using a function 'f(h, r)' that takes the hidden content 'h' and a random number 'r' as inputs. This ensures that different instances of the same hidden content are associated with different randomized messages, minimizing the correlation.
Encryption: The randomized message 'u' is then encrypted using a symmetric encryption scheme and a shared key 'K' distributed among a trusted group of users using a modified Diffie-Hellman protocol. This allows only the trusted group to decode the hidden content while preventing correlation attacks.
The authors analyze the security of the key distribution protocol under the Computational Diffie-Hellman Hypothesis and show that the proposed method effectively reduces the correlation between the publicly shared content and the hidden sensitive information, making it resistant to correlation attacks.
The method is designed to protect privacy even against the owner of the social network platform, as the encryption and key distribution are independent of the platform's mechanisms.
Na inny język
z treści źródłowej
arxiv.org
Głębsze pytania