We study the problem of searching on data that is encrypted using a public key system. Consider user Bob who sends email to user Alice encrypted under Alice’s public key. An email gateway wants to test whether the email contains the keyword “urgent” so that it could route the email accordingly. Alice, on the other hand does not wish to give the gateway the ability to decrypt all her messages. We define and construct a mechanism that enables Alice to provide a key to the gateway that enables the gateway to test whether the word “urgent” is a keyword in the email without learning anything else about the email. We refer to this mechanism as Public Key Encryption with keyword Search. As another example, consider a mail server that stores various messages publicly encrypted for Alice by others. Using our mechanism Alice can send the mail server a key that will enable the server to identify all messages containing some specific keyword, but learn nothing else. We define the concept of public key encryption with keyword search and give several constructions.

https://link.springer.com/content/pdf/10.1007%2F978-3-540-24676-3_30.pdf

Public Key Encryption with Keyword Search

Machine learning techniques based on neural networks are achieving remarkable results in a wide variety of domains. Often, the training of models requires large, representative datasets, which may be crowdsourced and contain sensitive information. The models should not expose private information in these datasets. Addressing this goal, we develop new algorithmic techniques for learning and a refined analysis of privacy costs within the framework of differential privacy. Our implementation and experiments demonstrate that we can train deep neural networks with non-convex objectives, under a modest privacy budget, and at a manageable cost in software complexity, training efficiency, and model quality.

Deep Learning with Differential Privacy

In this paper we address the issue of privacy preserving data mining. Specifically, we consider a scenario in which two parties owning confidential databases wish to run a data mining algorithm on the union of their databases, without revealing any unnecessary information. Our work is motivated by the need both to protect privileged information and to enable its use for research or other purposes.

The above problem is a specific example of secure multi-party computation and, as such, can be solved using known generic protocols. However, data mining algorithms are typically complex and, furthermore, the input usually consists of massive data sets. The generic protocols in such a case are of no practical use and therefore more efficient protocols are required. We focus on the problem of decision tree learning with the popular ID3 algorithm. Our protocol is considerably more efficient than generic solutions and demands both very few rounds of communication and reasonable bandwidth.

https://link.springer.com/content/pdf/10.1007/s00145-001-0019-2.pdf

Privacy Preserving Data Mining

We quantitatively investigate how machine learning models leak information about the individual data records on which they were trained. We focus on the basic membership inference attack: given a data record and black-box access to a model, determine if the record was in the model's training dataset. To perform membership inference against a target model, we make adversarial use of machine learning and train our own inference model to recognize differences in the target model's predictions on the inputs that it trained on versus the inputs that it did not train on. We empirically evaluate our inference techniques on classification models trained by commercial "machine learning as a service" providers such as Google and Amazon. Using realistic datasets and classification tasks, including a hospital discharge dataset whose membership is sensitive from the privacy perspective, we show that these models can be vulnerable to membership inference attacks. We then investigate the factors that influence this leakage and evaluate mitigation strategies.

/pdf/membership-inference-attacks-against-machine-learning-models-2tzchejf91.pdf

Membership Inference Attacks Against Machine Learning Models

Usually, a proof of a theorem contains more knowledge than the mere fact that the theorem is true. For instance, to prove that a graph is Hamiltonian it suffices to exhibit a Hamiltonian tour in it; however, this seems to contain more knowledge than the single bit Hamiltonian/non-Hamiltonian.In this paper a computational complexity theory of the “knowledge” contained in a proof is developed. Zero-knowledge proofs are defined as those proofs that convey no additional knowledge other than the correctness of the proposition in question. Examples of zero-knowledge proof systems are given for the languages of quadratic residuosity and 'quadratic nonresiduosity. These are the first examples of zero-knowledge proofs for languages not known to be efficiently recognizable.

/pdf/the-knowledge-complexity-of-interactive-proof-systems-31apre0ecf.pdf

The knowledge complexity of interactive proof-systems

In this paper we introduce the concept of privacy preserving data mining. In our model, two parties owning confidential databases wish to run a data mining algorithm on the union of their databases, without revealing any unnecessary information. This problem has many practical and important applications, such as in medical research with confidential patient records. Data mining algorithms are usually complex, especially as the size of the input is measured in megabytes, if not gigabytes. A generic secure multi-party computation solution, based on evaluation of a circuit computing the algorithm on the entire input, is therefore of no practical use. We focus on the problem of decision tree learning and use ID3, a popular and widely used algorithm for this problem. We present a solution that is considerably more efficient than generic solutions. It demands very few rounds of communication and reasonable bandwidth. In our solution, each party performs by itself a computation of the same order as computing the ID3 algorithm for its own database. The results are then combined using efficient cryptographic protocols, whose overhead is only logarithmic in the number of transactions in the databases. We feel that our result is a substantial contribution, demonstrating that secure multi-party computation can be made practical, even for complex problems and large inputs.

Privacy preserving data mining

In this paper, we survey the basic paradigms and notions of secure mul- tiparty computation and discuss their relevance to the fleld of privacy-preserving data mining. In addition to reviewing deflnitions and constructions for secure mul- tiparty computation, we discuss the issue of e-ciency and demonstrate the di-cul- ties involved in constructing highly e-cient protocols. We also present common errors that are prevalent in the literature when secure multiparty computation techniques are applied to privacy-preserving data mining. Finally, we discuss the relationship between secure multiparty computation and privacy-preserving data mining, and show which problems it solves and which problems it does not.

/pdf/secure-multiparty-computation-for-privacy-preserving-data-5brnfc9ewo.pdf

Secure Multiparty Computation for Privacy-Preserving Data Mining

We present FairplayMP (for "Fairplay Multi-Party"), a system for secure multi-party computation. Secure computation is one of the great achievements of modern cryptography, enabling a set of untrusting parties to compute any function of their private inputs while revealing nothing but the result of the function. In a sense, FairplayMP lets the parties run a joint computation that emulates a trusted party which receives the inputs from the parties, computes the function, and privately informs the parties of their outputs. FairplayMP operates by receiving a high-level language description of a function and a configuration file describing the participating parties. The system compiles the function into a description as a Boolean circuit, and perform a distributed evaluation of the circuit while revealing nothing else. FairplayMP supplements the Fairplay system [16], which supported secure computation between two parties. The underlying protocol of FairplayMP is the Beaver-Micali-Rogaway (BMR) protocol which runs in a constant number of communication rounds (eight rounds in our implementation). We modified the BMR protocol in a novel way and considerably improved its performance by using the Ben-Or-Goldwasser-Wigderson (BGW) protocol for the purpose of constructing gate tables. We chose to use this protocol since we believe that the number of communication rounds is a major factor on the overall performance of the protocol. We conducted different experiments which measure the effect of different parameters on the performance of the system and demonstrate its scalability. (We can now tell, for example, that running a second-price auction between four bidders, using five computation players, takes about 8 seconds.)

/pdf/fairplaymp-a-system-for-secure-multi-party-computation-4a5unfitfc.pdf

FairplayMP: a system for secure multi-party computation

Cloud storage systems are becoming increasingly popular. A promising technology that keeps their cost down is deduplication, which stores only a single copy of repeating data. Client-side deduplication attempts to identify deduplication opportunities already at the client and save the bandwidth of uploading copies of existing files to the server. In this work we identify attacks that exploit client-side deduplication, allowing an attacker to gain access to arbitrary-size files of other users based on a very small hash signatures of these files. More specifically, an attacker who knows the hash signature of a file can convince the storage service that it owns that file, hence the server lets the attacker download the entire file. (In parallel to our work, a subset of these attacks were recently introduced in the wild with respect to the Dropbox file synchronization service.) To overcome such attacks, we introduce the notion of proofs-of-ownership (PoWs), which lets a client efficiently prove to a server that that the client holds a file, rather than just some short information about it. We formalize the concept of proof-of-ownership, under rigorous security definitions, and rigorous efficiency requirements of Petabyte scale storage systems. We then present solutions based on Merkle trees and specific encodings, and analyze their security. We implemented one variant of the scheme. Our performance measurements indicate that the scheme incurs only a small overhead compared to naive client-side deduplication.

/pdf/proofs-of-ownership-in-remote-storage-systems-53grbrpq9r.pdf

Proofs of ownership in remote storage systems

As the volume of data increases, so does the demand for online storage services, from simple backup services to cloud storage infrastructures. Although deduplication is most effective when applied across multiple users, cross-user deduplication has serious privacy implications. Some simple mechanisms can enable cross-user deduplication while greatly reducing the risk of data leakage. Cloud storage refers to scalable and elastic storage capabilities delivered as a service using Internet technologies with elastic provisioning and usebased pricing that doesn't penalize users for changing their storage consumption without notice.

Benny Pinkas

Papers

Privacy preserving data mining

Secure Multiparty Computation for Privacy-Preserving Data Mining

FairplayMP: a system for secure multi-party computation

Proofs of ownership in remote storage systems

Side Channels in Cloud Services: Deduplication in Cloud Storage