In today's data-driven world, the proliferation of publicly available information raises security concerns due to the information leakage (IL) problem. IL involves unintentionally exposing sensitive information to unauthorized parties via observable system information. Conventional statistical approaches rely on estimating mutual information (MI) between observable and secret information for detecting ILs, face challenges of the curse of dimensionality, convergence, computational complexity, and MI misestimation. Though effective, emerging supervised machine learning based approaches to detect ILs are limited to the binary system, sensitive information, and lacks a comprehensive framework. To address these limitations, we establish a theoretical framework using statistical learning theory and information theory to quantify and detect IL accurately. Using automated machine learning, we demonstrate that MI can be accurately estimated by approximating the typically unknown Bayes predictor's LoG-Loss and accuracy. Based on this, we show how MI can effectively be estimated to detect ILs. Our method performs superior to state-of-the-art baselines in an empirical study considering synthetic and real-world OpenSSL TLS server datasets.

In anomaly detection, a prominent task is to induce a model to identify anomalies learned solely based on normal data. Generally, one is interested in finding an anomaly detector that correctly identifies anomalies, i.e., data points that do not belong to the normal class, without raising too many false alarms. Which anomaly detector is best suited depends on the dataset at hand and thus needs to be tailored. The quality of an anomaly detector may be assessed via confusion-based metrics such as the Matthews correlation coefficient (MCC). However, since during training only normal data is available in a semi-supervised setting, such metrics are not accessible. To facilitate automated machine learning for anomaly detectors, we propose to employ meta-learning to predict MCC scores using the metrics that can be computed with normal data only and order anomaly detectors using the predicted scores for selection. First promising results can be obtained considering the hypervolume and the false positive rate as meta-features.

The application of convolutional neural networks (CNNs) to break cryptographic systems through hardware side-channels facilitated rapid and adaptable attacks on cryptographic systems like smart cards and Trusted Platform Modules (TPMs). However, current approaches rely on manually designed CNN architectures by domain experts, which are time-consuming and impractical for attacking new systems. To overcome this, recent research has delved into the use of neural architecture search (NAS) to discover appropriate CNN architectures automatically. This approach aims to alleviate the burden on human experts and facilitate more efficient exploration of new attack targets. However, these works only optimize the architecture using the secret key information from the attack dataset and explore limited search strategies with one-dimensional CNNs. In this work, we propose a fully black-box NAS approach that solely utilizes the profiling dataset for optimization. Through an extensive experimental parameter study, we investigate which choices for NAS, such as using 1-D or 2-D CNNs and various search strategies, produce the best results on 10 state-of-the-art datasets for identity leakage model. Our results demonstrate that applying the Random search strategy on 1-D inputs achieves a high success rate, enabling retrieval of the correct secret key using a single attack trace on two datasets. This combination matches the attack efficiency of fixed CNN architectures and outperforms them in 4 out of 10 datasets. Our experiments also emphasize the importance of repeated attack evaluations for ML-based solutions to avoid biased performance estimates.

Choice functions accept a set of alternatives as input and produce a preferred subset of these alternatives as output. We study the problem of learning such functions under conditions of contextdependence of preferences, which means that the preference in favor of a certain choice alternative may depend on what other options are also available. In spite of its practical relevance, this kind of context-dependence has received little attention in preference learning so far. We propose a suitable model based on context-dependent (latent) utility functions, thereby reducing the problem to the task of learning such utility functions. Practically, this comes with a number of challenges. For example, the set of alternatives provided as input to a choice function can be of any size, and the output of the function should not depend on the order in which the alternatives are presented. To meet these requirements, we propose two general approaches based on two representations of context-dependent utility functions, as well as instantiations in the form of appropriate end-to-end trainable neural network architectures. Moreover, to demonstrate the performance of both networks, we present extensive empirical evaluations on both synthetic and real-world datasets.

Due to the proliferation of a large amount of publicly available data, information leakage (IL) has become a major problem. IL occurs when secret (sensitive) information of a system is inadvertently disclosed to unauthorized parties through externally observable information. Standard statistical approaches estimate the mutual information between observable (input) and secret information (output), which tends to be a difficult problem for high-dimensional input. Current approaches based on (supervised) machine learning using the accuracy of predictive models on extracted system input and output have proven to be more effective in detecting these leakages. However, these approaches are domain-specific and fail to account for imbalance in the dataset. In this paper, we present a robust autonomous approach to detecting IL, which blends machine learning and statistical techniques, to overcome these shortcomings. We propose to use Fisher’s Exact Test (FET) on the evaluated confusion matrix, which inherently takes the imbalances in the dataset into account. As a use case, we consider the problem of detecting padding side-channels or ILs in systems implementing cryptographic protocols. In an extensive experimental study on detecting ILs in synthetic and real-world scenarios, our approach outperforms the state of the art.

Currently most practical attacks on cryptographic protocols like TLS are based on side channels, such as padding oracles. Some well-known recent examples are DROWN, ROBOT and Raccoon (USENIX Security 2016, 2018, 2021). Such attacks are usually found by careful and time-consuming manual analysis by specialists. In this paper, we consider the question of how such attacks can be systematically detected and prevented before (large-scale) deployment. We propose a new, fully automated approach, which uses supervised learning to identify arbitrary patterns in network protocol traffic. In contrast to classical scanners, which search for known side channels, the detection of general patterns might detect new side channels, even unexpected ones, such as those from the ROBOT attack. To analyze this approach, we develop a tool to detect Bleichenbacher-like padding oracles in TLS server implementations, based on an ensemble of machine learning algorithms. We verify that the approach indeed detects known vulnerabilities successfully and reliably. The tool also provides detailed information about detected patterns to developers, to assist in removing a potential padding oracle. Due to the automation, the approach scales much better than manual analysis and could even be integrated with a CI/CD pipeline of a development environment, for example.

We study the problem of learning choice functions, which play an important role in various domains of application, most notably in the field of economics. Formally, a choice function is a mapping from sets to sets: Given a set of choice alternatives as input, a choice function identifies a subset of most preferred elements. Learning choice functions from suitable training data comes with a number of challenges. For example, the sets provided as input and the subsets produced as output can be of any size. Moreover, since the order in which alternatives are presented is irrelevant, a choice function should be symmetric. Perhaps most importantly, choice functions are naturally context-dependent, in the sense that the preference in favor of an alternative may depend on what other options are available. We formalize the problem of learning choice functions and present two general approaches based on two representations of context-dependent utility functions. Both approaches are instantiated by means of appropriate neural network architectures, and their performance is demonstrated on suitable benchmark tasks.

Object ranking is an important problem in the realm of preference learning. On the basis of training data in the form of a set of rankings of objects, which are typically represented as feature vectors, the goal is to learn a ranking function that predicts a linear order of any new set of objects. Current approaches commonly focus on ranking by scoring, i.e., on learning an underlying latent utility function that seeks to capture the inherent utility of each object. These approaches, however, are not able to take possible effects of context-dependence into account, where context-dependence means that the utility or usefulness of an object may also depend on what other objects are available as alternatives. In this paper, we formalize the problem of context-dependent ranking and present two general approaches based on two natural representations of context-dependent ranking functions. Both approaches are instantiated by means of appropriate neural network architectures, which are evaluated on suitable benchmark task.

Schedae Informaticae » 2016 » Volume 25 » Pairwise versus Pointwise Ranking: A Case Study A A A