In this paper, we shed new light on the DNS amplification ecosystem, by studying complementary data sources, bolstered by orthogonal methodologies. First, we introduce a passive attack detection method for the Internet core, i.e., at Internet eXchange Points (IXPs). Surprisingly, IXPs and honeypots observe mostly disjoint sets of attacks: 96% of IXP-inferred attacks were invisible to a sizable honeypot platform. Second, we assess the effectiveness of observed DNS attacks by studying IXP traces jointly with diverse data from independent measurement infrastructures. We find that attackers efficiently detect new reflectors and purposefully rotate between them. At the same time, we reveal that attackers are a small step away from bringing about significantly higher amplification factors (14x). Third, we identify and fingerprint a major attack entity by studying patterns in attack traces. We show that this entity dominates the DNS amplification ecosystem by carrying out 59% of the attacks, and provide an in-depth analysis of its behavior over time. Finally, our results reveal that operators of various .gov names do not adhere to DNSSEC key rollover best practices, which exacerbates amplification potential. We can verifiably connect this operational behavior to misuses and attacker decision-making.
相關內容
With the increasingly rapid development of new malicious computer software by bad faith actors, both commercial and research-oriented antivirus detectors have come to make greater use of machine learning tactics to identify such malware as harmful before end users are exposed to their effects. This, in turn, has spurred the development of tools that allow for known malware to be manipulated such that they can evade being classified as dangerous by these machine learning-based detectors, while retaining their malicious functionality. These manipulations function by applying a set of changes that can be made to Windows programs that result in a different file structure and signature without altering the software's capabilities. Various proposals have been made for the most effective way of applying these alterations to input malware to deceive static malware detectors; the purpose of this research is to examine these proposals and test their implementations to determine which tactics tend to generate the most successful attacks.
COVID-19 vaccines have been rolled out in many countries and with them a number of vaccination certificates. For instance, the EU is utilizing a digital certificate in the form of a QR-code that is digitally signed and can be easily validated throughout all EU countries. In this paper, we investigate the current state of the COVID-19 vaccination certificate market in the darkweb with a focus on the EU Digital Green Certificate (DGC). We investigate $17$ marketplaces and $10$ vendor shops, that include vaccination certificates in their listings. Our results suggest that a multitude of sellers in both types of platforms are advertising selling capabilities. According to their claims, it is possible to buy fake vaccination certificates issued in most countries worldwide. We demonstrate some examples of such sellers, including how they advertise their capabilities, and the methods they claim to be using to provide their services. We highlight two particular cases of vendor shops, with one of them showing an elevated degree of professionalism, showcasing forged valid certificates, the validity of which we verify using two different national mobile COVID-19 applications.
Data privacy is critical in instilling trust and empowering the societal pacts of modern technology-driven democracies. Unfortunately, it is under continuous attack by overreaching or outright oppressive governments, including some of the world's oldest democracies. Increasingly-intrusive anti-encryption laws severely limit the ability of standard encryption to protect privacy. New defense mechanisms are needed. Plausible deniability (PD) is a powerful property, enabling users to hide the existence of sensitive information in a system under direct inspection by adversaries. Popular encrypted storage systems such as TrueCrypt and other research efforts have attempted to also provide plausible deniability. Unfortunately, these efforts have often operated under less well-defined assumptions and adversarial models. Careful analyses often uncover not only high overheads but also outright security compromise. Further, our understanding of adversaries, the underlying storage technologies, as well as the available plausible deniable solutions have evolved dramatically in the past two decades. The main goal of this work is to systematize this knowledge. It aims to: - identify key PD properties, requirements, and approaches; - present a direly-needed unified framework for evaluating security and performance; - explore the challenges arising from the critical interplay between PD and modern system layered stacks; - propose a new "trace-oriented" PD paradigm, able to decouple security guarantees from the underlying systems and thus ensure a higher level of flexibility and security independent of the technology stack. This work is meant also as a trusted guide for system and security practitioners around the major challenges in understanding, designing, and implementing plausible deniability into new or existing systems.
Graph neural networks, a popular class of models effective in a wide range of graph-based learning tasks, have been shown to be vulnerable to adversarial attacks. While the majority of the literature focuses on such vulnerability in node-level classification tasks, little effort has been dedicated to analysing adversarial attacks on graph-level classification, an important problem with numerous real-life applications such as biochemistry and social network analysis. The few existing methods often require unrealistic setups, such as access to internal information of the victim models, or an impractically-large number of queries. We present a novel Bayesian optimisation-based attack method for graph classification models. Our method is black-box, query-efficient and parsimonious with respect to the perturbation applied. We empirically validate the effectiveness and flexibility of the proposed method on a wide range of graph classification tasks involving varying graph properties, constraints and modes of attack. Finally, we analyse common interpretable patterns behind the adversarial samples produced, which may shed further light on the adversarial robustness of graph classification models.
Feature attribution is often loosely presented as the process of selecting a subset of relevant features as a rationale of a prediction. This lack of clarity stems from the fact that we usually do not have access to any notion of ground-truth attribution and from a more general debate on what good interpretations are. In this paper we propose to formalise feature selection/attribution based on the concept of relaxed functional dependence. In particular, we extend our notions to the instance-wise setting and derive necessary properties for candidate selection solutions, while leaving room for task-dependence. By computing ground-truth attributions on synthetic datasets, we evaluate many state-of-the-art attribution methods and show that, even when optimised, some fail to verify the proposed properties and provide wrong solutions.
A key challenge of big data analytics is how to collect a large volume of (labeled) data. Crowdsourcing aims to address this challenge via aggregating and estimating high-quality data (e.g., sentiment label for text) from pervasive clients/users. Existing studies on crowdsourcing focus on designing new methods to improve the aggregated data quality from unreliable/noisy clients. However, the security aspects of such crowdsourcing systems remain under-explored to date. We aim to bridge this gap in this work. Specifically, we show that crowdsourcing is vulnerable to data poisoning attacks, in which malicious clients provide carefully crafted data to corrupt the aggregated data. We formulate our proposed data poisoning attacks as an optimization problem that maximizes the error of the aggregated data. Our evaluation results on one synthetic and two real-world benchmark datasets demonstrate that the proposed attacks can substantially increase the estimation errors of the aggregated data. We also propose two defenses to reduce the impact of malicious clients. Our empirical results show that the proposed defenses can substantially reduce the estimation errors of the data poisoning attacks.
Training machine learning models in a meaningful order, from the easy samples to the hard ones, using curriculum learning can provide performance improvements over the standard training approach based on random data shuffling, without any additional computational costs. Curriculum learning strategies have been successfully employed in all areas of machine learning, in a wide range of tasks. However, the necessity of finding a way to rank the samples from easy to hard, as well as the right pacing function for introducing more difficult data can limit the usage of the curriculum approaches. In this survey, we show how these limits have been tackled in the literature, and we present different curriculum learning instantiations for various tasks in machine learning. We construct a multi-perspective taxonomy of curriculum learning approaches by hand, considering various classification criteria. We further build a hierarchical tree of curriculum learning methods using an agglomerative clustering algorithm, linking the discovered clusters with our taxonomy. At the end, we provide some interesting directions for future work.
Recently, NLP has seen a surge in the usage of large pre-trained models. Users download weights of models pre-trained on large datasets, then fine-tune the weights on a task of their choice. This raises the question of whether downloading untrusted pre-trained weights can pose a security threat. In this paper, we show that it is possible to construct ``weight poisoning'' attacks where pre-trained weights are injected with vulnerabilities that expose ``backdoors'' after fine-tuning, enabling the attacker to manipulate the model prediction simply by injecting an arbitrary keyword. We show that by applying a regularization method, which we call RIPPLe, and an initialization procedure, which we call Embedding Surgery, such attacks are possible even with limited knowledge of the dataset and fine-tuning procedure. Our experiments on sentiment classification, toxicity detection, and spam detection show that this attack is widely applicable and poses a serious threat. Finally, we outline practical defenses against such attacks. Code to reproduce our experiments is available at //github.com/neulab/RIPPLe.
In recent years with the rise of Cloud Computing (CC), many companies providing services in the cloud, are empowered a new series of services to their catalog, such as data mining (DM) and data processing, taking advantage of the vast computing resources available to them. Different service definition proposals have been proposed to address the problem of describing services in CC in a comprehensive way. Bearing in mind that each provider has its own definition of the logic of its services, and specifically of DM services, it should be pointed out that the possibility of describing services in a flexible way between providers is fundamental in order to maintain the usability and portability of this type of CC services. The use of semantic technologies based on the proposal offered by Linked Data (LD) for the definition of services, allows the design and modelling of DM services, achieving a high degree of interoperability. In this article a schema for the definition of DM services on CC is presented, in addition are considered all key aspects of service in CC, such as prices, interfaces, Software Level Agreement, instances or workflow of experimentation, among others. The proposal presented is based on LD, so that it reuses other schemata obtaining a best definition of the service. For the validation of the schema, a series of DM services have been created where some of the best known algorithms such as \textit{Random Forest} or \textit{KMeans} are modeled as services.
Embedding network data into a low-dimensional vector space has shown promising performance for many real-world applications, such as node classification and entity retrieval. However, most existing methods focused only on leveraging network structure. For social networks, besides the network structure, there also exists rich information about social actors, such as user profiles of friendship networks and textual content of citation networks. These rich attribute information of social actors reveal the homophily effect, exerting huge impacts on the formation of social networks. In this paper, we explore the rich evidence source of attributes in social networks to improve network embedding. We propose a generic Social Network Embedding framework (SNE), which learns representations for social actors (i.e., nodes) by preserving both the structural proximity and attribute proximity. While the structural proximity captures the global network structure, the attribute proximity accounts for the homophily effect. To justify our proposal, we conduct extensive experiments on four real-world social networks. Compared to the state-of-the-art network embedding approaches, SNE can learn more informative representations, achieving substantial gains on the tasks of link prediction and node classification. Specifically, SNE significantly outperforms node2vec with an 8.2% relative improvement on the link prediction task, and a 12.7% gain on the node classification task.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191