Privacy is a fundamental challenge for a variety of smart applications that depend on data aggregation and collaborative learning across different entities. In this paper, we propose a novel privacy-preserved architecture where clients can collaboratively train a deep model while preserving the privacy of each client’s data. Our main strategy is to carefully partition a deep neural network to two non-colluding parties. One party performs linear computations on encrypted data utilizing a less complex homomorphic cryptosystem, while the other executes non-polynomial computations in plaintext but in a privacy-preserved manner. We analyze security and compare the communication and computation complexity with the existing approaches. Our extensive experiments on different datasets demonstrate not only stable training without accuracy loss, but also 14 to 35 times speedup compared to the state-of-the-art system.
— In this paper, for the first time, we define a general\ud
notion for proxy re-encryption (PRE), which we call deterministic\ud
finite automata-based functional PRE (DFA-based FPRE).\ud
Meanwhile, we propose the first and concrete DFA-based FPRE\ud
system, which adapts to our new notion. In our scheme, a message\ud
is encrypted in a ciphertext associated with an arbitrary length\ud
index string, and a decryptor is legitimate if and only if a DFA\ud
associated with his/her secret key accepts the string. Furthermore,\ud
the above encryption is allowed to be transformed to another\ud
ciphertext associated with a new string by a semitrusted proxy\ud
to whom a re-encryption key is given. Nevertheless, the proxy\ud
cannot gain access to the underlying plaintext. This new primitive\ud
can increase the flexibility of users to delegate their decryption\ud
rights to others. We also prove it as fully chosen-ciphertext secure\ud
in the standard model
Attribute Based Broadcast Encryption (ABBE) is a combination of Attribute Based Encryption (ABE) and Broadcast Encryption (BE). It allows a broadcaster (or encrypter) to broadcast an encrypted message that can only be decrypted by the receivers who are within a predefined user set and satisfy the access policy specified by the broadcaster. Compared with normal ABE, ABBE allows direct revocation, which is important in many real-time broadcasting applications such as Pay TV. In this paper, we propose two novel ABBE schemes that have distinguishing features: the first scheme is key-policy based and has short ciphertext and constant size decryption key; and the second one is ciphertext-policy based and has constant size ciphertext and short decryption key. Both of our schemes allow access policies to be expressed using AND-gate with positive, negative, and wildcard symbols, and are proven secure under the Decision n-BDHE assumption without random oracles.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.