We propose a novel end-to-end privacy-preserving framework, instantiated by three efficient protocols for different deployment scenarios, covering both input and output privacy, for the vertically split scenario in federated learning (FL), where features are split across clients and labels are not shared by all parties. We do so by distributing the role of the aggregator in FL into multiple servers and having them run secure multiparty computation (MPC) protocols to perform model and feature aggregation and apply differential privacy (DP) to the final released model. While a naive solution would have the clients delegating the entirety of training to run in MPC between the servers, our optimized solution, which supports purely global and also global-local models updates with privacy-preserving, drastically reduces the amount of computation and communication performed using multiparty computation. The experimental results also show the effectiveness of our protocols.

We propose a privacy-preserving Naive Bayes classifier and apply it to the problem of private text classification. In this setting, a party (Alice) holds a text message, while another party (Bob) holds a classifier. At the end of the protocol, Alice will only learn the result of the classifier applied to her text input and Bob learns nothing. Our solution is based on Secure Multiparty Computation (SMC). Our Rust implementation provides a fast and secure solution for the classification of unstructured text. Applying our solution to the case of spam detection (the solution is generic, and can be used in any other scenario in which the Naive Bayes classifier can be employed), we can classify an SMS as spam or ham in less than 340ms in the case where the dictionary size of Bob’s model includes all words ($n = 5200$) and Alice’s SMS has at most$m = 160$unigrams. In the case with$n = 369$and$m = 8$(the average of a spam SMS in the database), our solution takes only 21ms.

Noisy channels are a valuable resource from a cryptographic point of view. They can be used for exchanging secret-keys as well as realizing other cryptographic primitives such as commitment and oblivious transfer. To be really useful, noisy channels have to be considered in the scenario where a cheating party has some degree of control over the channel characteristics. Damgård et al. (EUROCRYPT 1999) proposed a more realistic model where such level of control is permitted to an adversary, the so called unfair noisy channels, and proved that they can be used to obtain commitment and oblivious transfer protocols. Given that noisy channels are a precious resource for cryptographic purposes, one important question is determining the optimal rate in which they can be used. The commitment capacity has already been determined for the cases of discrete memoryless channels and Gaussian channels. In this work we address the problem of determining the commitment capacity of unfair noisy channels. We compute a single-letter characterization of the commitment capacity of unfair noisy channels. In the case where an adversary has no control over the channel (the fair case) our capacity reduces to the well-known capacity of a discrete memoryless binary symmetric channel.

Domain Generation Algorithms (DGAs) are used by malware to generate a deterministic set of domains, usually by utilizing a pseudo-random seed. A malicious botmaster can establish connections between their command-and-control center (C&C) and any malware-infected machines by registering domains that will be DGA-generated given a specific seed, rendering traditional domain blacklisting ineffective. Given the nature of this threat, the real-time detection of DGA domains based on incoming DNS traffic is highly important. The use of neural network machine learning (ML) models for this task has been well-studied, but there is still substantial room for improvement. In this paper, we propose to use Inductive Venn-Abers predictors (IVAPs) to calibrate the output of existing ML models for DGA classification. The IVAP is a computationally efficient procedure which consistently improves the predictive accuracy of classifiers at the expense of not offering predictions for a small subset of inputs and consuming an additional amount of training data.

As the emergence of the Internet of Things (IoT) and the growing number of IoT devices, a stable connection service has become one of the key factors concerning the Quality of Service (QoS) provision. How to anticipate the peak traffic volume is essential. If the resource allocation is under provisioned, the service becomes susceptible to failure or security breach. Unfortunately, peak volumes are not captured in the systematic components of data and as a result conventional trend prediction methods have proven insufficient. We propose a framework that implements neural networks with filtering and a cost-adaptive loss function to improve the ability to predict peak volumes. Implementing this method on a real Domain Name Server (DNS) traffic data, we observe not only the improvement in the prediction performance but also a shorter lag time to predict peak values, which demonstrates our proposed method.

Classification of personal text messages has many useful applications in surveillance, e-commerce, and mental health care, to name a few. Giving applications access to personal texts can easily lead to (un)intentional privacy violations. We propose the first privacy-preserving solution for text classification that is provably secure. Our method, which is based on Secure Multiparty Computation (SMC), encompasses both feature extraction from texts, and subsequent classification with logistic regression and tree ensembles. We prove that when using our secure text classification method, the application does not learn anything about the text, and the author of the text does not learn anything about the text classification model used by the application beyond what is given by the classification result itself. We perform end-to-end experiments with an application for detecting hate speech against women and immigrants, demonstrating excellent runtime results without loss of accuracy.

Many data-driven personalized services require that private data of users is scored against a trained machine learning model. In this paper we propose a novel protocol for privacy-preserving classification of decision trees, a popular machine learning model in these scenarios. Our solutions is composed out of building blocks, namely a secure comparison protocol, a protocol for obliviously selecting inputs, and a protocol for multiplication. By combining some of the building blocks for our decision tree classification protocol, we also improve previously proposed solutions for classification of support vector machines and logistic regression models. Our protocols are information theoretically secure and, unlike previously proposed solutions, do not require modular exponentiations. We show that our protocols for privacy-preserving classification lead to more efficient results from the point of view of computational and communication complexities. We present accuracy and runtime results for seven classification benchmark datasets from the UCI repository.

Many machine learning (ML) applications rely on large amounts of personal data for training and inference. Among the most intimate exploited data sources is electroencephalogram (EEG) data. The emergence of consumer -grade, low-cost brain -computer interfaces (BCIs) and corresponding software development kits' is bringing the use of BCI within reach of application developers. The access that BCI applications have to neural signals rightly raises privacy concerns. Application developers can easily gain knowledge beyond the professed scope from unprotected EEG signals, including passwords, ATM PINs, and other personal data. The challenge is how to engage in meaningful ML with EEG data while protecting the privacy of users.

The content generated by users on social media is rich in personal information that can be mined to construct accurate user profiles, and subsequently used for tailored advertising or other personalized services. Facebook has recently come under scrutiny after a third party gained access to the data of millions of users and mined it to construct psychographical profiles, which were allegedly used to influence voters in elections. As part of a possible solution to avoid data breaches while still being able to perform meaningful machine learning (ML) on social media data, we propose a privacy-preserving algorithm for k-nearest neighbor (kNN) [1] , one of the oldest ML methods, used traditionally in collaborative filtering recommender systems.

Machine Learning (ML) techniques now impact a wide variety of domains. Highly regulated industries such as healthcare and finance have stringent compliance and data governance policies around data sharing. Advances in secure multiparty computation (SMC) for privacy-preserving machine learning (PPML) can help transform these regulated industries by allowing ML computations over encrypted data with personally identifiable information (PII). Yet very little of SMC-based PPML has been put into practice so far. In this paper we present the very first framework for privacy-preserving classification of tree ensembles with application in healthcare. We first describe the underlying cryptographic protocols that enable a healthcare organization to send encrypted data securely to a ML scoring service and obtain encrypted class labels without the scoring service actually seeing that input in the clear. We then describe the deployment challenges we solved to integrate these protocols in a cloud based scalable risk-prediction platform with multiple ML models for healthcare AI. Included are system internals, and evaluations of our deployment for supporting physicians to drive better clinical outcomes in an accurate, scalable, and provably secure manner. To the best of our knowledge, this is the first such applied framework with SMC-based privacy-preserving machine learning for healthcare.

Domain Generation Algorithms (DGAs) are a popular technique used by contemporary malware for command-and-control (C&C) purposes. Such malware utilizes DGAs to create a set of domain names that, when resolved, provide information necessary to establish a link to a C&C server. Automated discovery of such domain names in real-time DNS traffic is critical for network security as it allows to detect infection, and, in some cases, take countermeasures to disrupt the communication and identify infected machines. Detection of the specific DGA malware family provides the administrator valuable information about the kind of infection and steps that need to be taken. In this paper we compare and evaluate machine learning methods that classify domain names as benign or DGA, and label the latter according to their malware family. Unlike previous work, we select data for test and training sets according to observation time and known seeds. This allows us to assess the robustness of the trained classifiers for detecting domains generated by the same families at a different time or when seeds change. Our study includes tree ensemble models based on human-engineered features and deep neural networks that learn features automatically from domain names. We find that all state-of-the-art classifiers are significantly better at catching domain names from malware families with a time-dependent seed compared to time-invariant DGAs. In addition, when applying the trained classifiers on a day of real traffic, we find that many domain names unjustifiably are flagged as malicious, thereby revealing the shortcomings of relying on a standard whitelist for training a production grade DGA detection system.

Recently several different deep learning architectures have been proposed that take a string of characters as the raw input signal and automatically derive features for text classification. Few studies are available that compare the effectiveness of these approaches for character based text classification with each other. In this paper we perform such an empirical comparison for the important cybersecurity problem of DGA detection: classifying domain names as either benign vs. produced by malware (i.e., by a Domain Generation Algorithm). Training and evaluating on a dataset with 2M domain names shows that there is surprisingly little difference between various convolutional neural network (CNN) and recurrent neural network (RNN) based architectures in terms of accuracy, prompting a preference for the simpler architectures, since they are faster to train and to score, and less prone to overfitting.

The bounded storage model restricts the memory of an adversary in a cryptographic protocol, rather than restricting its computational power, making information theoretically secure protocols feasible. We present the first protocols for commitment and oblivious transfer in the bounded storage model with errors, i.e., the model where the public random sources available to the two parties are not exactly the same, but instead are only required to have a small Hamming distance between themselves. Commitment and oblivious transfer protocols were known previously only for the error-free variant of the bounded storage model, which is harder to realize.

Summary Despite the evident benefits of the access to virtually unlimited computational resources in cloud environments, enterprises and researchers still face upending challenges when deploying applications that deal with sensitive information to the cloud. That is specially true for medical or tax records, for which there are strong legal restrictions to data escrow. In these cases one must be certain that a third party, such as the cloud provider, will never have access to the data. This work presents a solid access control framework that uses hybrid cryptography at client‐side and a two‐factor authentication technique to guarantee a secure key management protocol. We also demonstrate the use of homomorphic and order‐preserving encryption as a viable solution for the computation of regular searches over electronic health records in the cloud, while preserving the confidentiality of clinical data and the privacy of patients, even in the face of a semi‐honest, or “honest, but curious,” cloud provider. We introduce a trusted element, a browser extension, to prevent attacks from malicious cloud providers. The result is evaluated through a full‐featured prototype that manages health records modeled with a few OpenEHR archetypes. The prototype can be easily extended to handle any data structure modeled with OpenEHR.

Noisy channels are a powerful resource for cryptography as they can be used to obtain information-theoretic secure key agreement, commitment, and oblivious transfer protocols, among others. Oblivious transfer (OT) is a fundamental primitive, since it is complete for secure multi-party computation, and the OT capacity characterizes how efficiently a channel can be used for obtaining string oblivious transfer. Ahlswede and Csiszár (ISIT'07) presented upper and lower bounds on the OT capacity of generalized erasure channels (GECs) against passive adversaries. In the case of GEC with erasure probability at least 1/2, the upper and lower bounds match and, therefore, the OT capacity was determined. It was later proved by Pinto et al. [IEEE Trans. Inf. Theory 57(8)] that the OT capacity is identical for passive and malicious adversaries. In the case of GEC with erasure probability smaller than 1/2, the known lower bound against passive adversaries that was established by Ahlswede and Csiszár does not match their upper bound and it was unknown whether this OT rate could be achieved against malicious adversaries as well. In this paper, we show that there is a protocol against malicious adversaries achieving the same OT rate that was obtained against passive adversaries. We obtain our results by a new combination of interactive hashing and typicality tests that are suitable for dealing with the case of low erasure probability (p* <;1/2 ).

User profiling from user generated content (UGC) is a common practice that supports the business models of many social media companies. Existing systems require that the UGC is fully exposed to the module that constructs the user profiles. In this paper we show that it is possible to build user profiles without ever accessing the user's original data, and without exposing the trained machine learning models for user profiling - which are the intellectual property of the company - to the users of the social media site. We present VirtualIdentity, an application that uses secure multi-party cryptographic protocols to detect the age, gender and personality traits of users by classifying their user-generated text and personal pictures with trained support vector machine models in a privacy preserving manner.

Linear algebra operations on private distributed data are frequently required in several practical scenarios (e.g., statistical analysis and privacy preserving databases). We present universally composable two-party protocols to compute inner products, determinants, eigenvalues, and eigenvectors. These protocols are built for a two-party scenario where the inputs are provided by mutually distrustful parties. After execution, the protocols yield the results of the intended operation while preserving the privacy of their inputs. Universal composability is obtained in the trusted initializer model, ensuring information theoretical security under arbitrary protocol composition in complex environments. Furthermore, our protocols are computationally efficient since they only require field multiplication and addition operations.

Governments can take advantage of cloud computing to deal with the pressures as well as challenges to reduce costs and improve service performance. In this context, the joint evaluation of performance and dependability (i.e., Performability) is very important, and stochastic modeling provides a prominent mechanism for assessing government services in the cloud and evaluation of distinct scenarios. This work proposes an approach based on stochastic Petri nets (SPN) for performability evaluation of Taxpayer Assistance Center (TAC) in the cloud. Performability techniques are considered, as failures in the process are taken into account and the respective impact in performance is analyzed. Results are presented to demonstrate the feasibility of the proposed approach, and they indicate that failures should not be neglected.

In Cramer et al. (2007, Bounded CCA2-Secure Encryption. In Kurosawa, K. (ed.), Advances in Cryptology – ASIACRYPT 2007, Kuching, Malaysia, December 2–6, Lecture Notes in Computer Science, Vol. 4833, pp. 502–518. Springer, Berlin, Germany) proposed a public-key encryption scheme secure against adversaries with a bounded number of decryption queries based on the decisional Diffie–Hellman problem. In this paper, we show that the same result can be obtained based on weaker computational assumptions, namely: the computational Diffie–Hellman and the hashed Diffie–Hellman assumptions.

In the bounded storage model the memory of the adversarial parties is restricted, instead of their computational power. This different restriction allows the construction of protocols with information-theoretical (instead of only computational) security. We present the first protocol for oblivious transfer in the bounded storage model with errors, i.e., where the public random sources available to the two parties are not exactly the same, but instead are only required to have a small Hamming distance between themselves, and the memory of the (adversarial) receiver is limited. Oblivious transfer protocols were known previously only for the error-free variant of the bounded storage model, which is harder to realize.

Over the years, inbound logistics has been prominent to many industries, as globalization has forced companies to increase productivity while reducing costs. Transport modes considerably impact the costs, and a proper evaluation should consider performance, failures as well as sustainability issues. This paper presents an approach based on stochastic Petri nets (SPN) for assessing different transport modes in inbound logistics for coffee industry taking into account performability and sustainability.

The McEliece public-key encryption scheme has become an interesting alternative to cryptosystems based on number-theoretical problems. Different from RSA and ElGamal, McEliece PKC is not known to be broken by a quantum computer. Moreover, even though McEliece PKC has a relatively big key size, encryption and decryption operations are rather efficient. In spite of all the recent results in coding-theory-based cryptosystems, to the date, there are no constructions secure against chosen ciphertext attacks in the standard model-the de facto security notion for public-key cryptosystems. In this paper, we show the first construction of a McEliece-based public-key cryptosystem secure against chosen ciphertext attacks in the standard model. Our construction is inspired by a recently proposed technique by Rosen and Segev.

We introduce the first efficient fully simulatable construction of oblivious transfer based on the McEliece assumptions in the common reference string model.

Information-theoretically secure string oblivious transfer (OT) can be constructed based on discrete memoryless channel (DMC). The oblivious transfer capacity of a channel characterizes - similarly to the (standard) information capacity - how efficiently it can be exploited for secure oblivious transfer of strings. The OT capacity of a generalized erasure channel (GEC) - which is a combination of a (general) DMC with the erasure channel - has been established by Ahlswede and Csizar at ISIT'07 in the case of passive adversaries. In this paper, we present the protocol that achieves this capacity against malicious adversaries for GEC with erasure probability at least 1/2. Our construction is based on the protocol of Crepeau and Savvides from Eurocrypt'06 which uses interactive hashing (IH). We solve an open question posed by the above paper, by basing it upon a constant round IH scheme (previously proposed by Ding et al. at TCC'04). As a side result, we show that the Ding et al. IH protocol can deal with transmission errors.

We prove that the commitment capacity of the power-constrained Gaussian channel, i.e., the optimal rate at which this channel can be used for implementing commitment schemes, is infinite.

In this paper, we deal with the task of obtaining oblivious transfer (OT) from noisy resources. We characterize which noisy channels/distributions are useful for obtaining OT. We also introduce the problem of computing the oblivious-transfer capacity of a noisy resource, which measures the optimal way of implementing OT from a noisy channel/distribution. We show that for honest-but-curious sender, the oblivious-transfer capacity of noisy resources is strictly positive. Several open questions are raised.

We consider bit commitment over additive white Gaussian noise channels. Our main result is that the maximum rate at which this class of channels can be used for implementing commitment protocols (the commitment capacity) is provably infinite, even under an average power constraint.

One of the most important primitives in two-party distrustful cryptography is oblivious transfer, a complete primitive for two-party computation. Recently introduced, the oblivious transfer capacity of a noisy channel measures an efficiency of information theoretical reductions from 1-out-of-k, l-string oblivious transfer to noisy channels. It is defined as the maximal achievable ratio l/n, where l is the length of the strings which are to be transferred and n is the number of times the noisy channel is invoked. This quantity is unknown in a general case. For discrete memoryless channels, it is known to be nonnegligible for honest-but-curious players, but the non-zero rates have not ever been proved achievable in the case of malicious players. Here, we show that in the particular case of the erasure channel, more precise answers can be obtained. We compute the OT capacity of the erasure channel for the case of honest-but-curious players and, for the fully malicious players, we give its lower bound.

Bit commitment is an important tool for constructing zero-knowledge proofs and multi-party computation. Unconditionally secure bit commitment can be based, in particular, on noisy channel or correlation where noise considered a valuable resource. Recently, Winter, Nascimento and Imai introduced the concept of commitment capacity, the maximal ratio between the length of a string which the sender commits to and the number of times the noisy channel/correlation is used. They also proved that for any discrete memoryless channel there exists a secure protocol achieving its commitment capacity however, no particular construction was given. Solving their open question, we provide an efficient protocol for achieving the commitment capacity of discrete memoryless systems (noisy channels and correlations).

We deal with the task of obtaining oblivious transfer from noisy resources. We characterize which noisy channels/distributions are useful for obtaining oblivious transfer. We also introduce the problem of computing the oblivious transfer capacity of a noisy resource, which measures the optimal way of implementing oblivious transfer from a noisy channel/distribution. We show that for honest but curious sender, the oblivious transfer capacity of noisy resources is strictly positive. Several open questions are raised.

This paper studies the optimisation of the channel with cryptographic primitives such as coin tossing and oblivious transfer by committing to a set of strings. The main contribution of this paper is that the commitment is possible from any nontrivial correlation at rates when the sender is Alice and Bob, those rates are optimal. Also the coin tossing capacity is infinite for every channel having a positive bit commitment rate.

Summary form only given. We discuss information-theoretic methods to prove the security of cryptosystems. We study what is called, unconditionally secure (or information-theoretically secure) cryptographic schemes in search for a system that can provide long-term security and that does not impose limits on the adversary's computational power.