Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 9
Encontrada(s) 1 página(s)

Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

LSGAN-AT: enhancing malware detector robustness against adversarial examples

  • Wang, J.
  • Chang, X.
  • Wang, Y.
  • Rodríguez, R.J.
  • Zhang, J.
Adversarial Malware Example (AME)-based adversarial training can effectively enhance the robustness of Machine Learning (ML)-based malware detectors against AME. AME quality is a key factor to the robustness enhancement. Generative Adversarial Network (GAN) is a kind of AME generation method, but the existing GAN-based AME generation methods have the issues of inadequate optimization, mode collapse and training instability. In this paper, we propose a novel approach (denote as LSGAN-AT) to enhance ML-based malware detector robustness against Adversarial Examples, which includes LSGAN module and AT module. LSGAN module can generate more effective and smoother AME by utilizing brand-new network structures and Least Square (LS) loss to optimize boundary samples. AT module makes adversarial training using AME generated by LSGAN to generate ML-based Robust Malware Detector (RMD). Extensive experiment results validate the better transferability of AME in terms of attacking 6 ML detectors and the RMD transferability in terms of resisting the MalGAN black-box attack. The results also verify the performance of the generated RMD in the recognition rate of AME. © 2021, The Author(s).
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Defense and Attack Techniques against File-based TOCTOU Vulnerabilities: a Systematic Review

  • Raducu, R.
  • Rodríguez Fernández, R. J.
  • Álvarez Pérez-Aradros, P.
File-based Time-of-Check to Time-of-Use (TOCTOU) race conditions are a well-known type of security vulnerability. A wide variety of techniques have been proposed to detect, mitigate, avoid, and exploit these vulnerabilities over the past 35 years. However, despite these research efforts, TOCTOU vulnerabilities remain unsolved due to their non-deterministic nature and the particularities of the different filesystems involved in running vulnerable programs, especially in Unix-like operating system environments. In this paper, we present a systematic literature review on defense and attack techniques related to the file-based TOCTOU vulnerability. We apply a reproducible methodology to search, filter, and analyze the most relevant research proposals to define a global and understandable vision of existing solutions. The results of this analysis are finally used to discuss future research directions that can be explored to move towards a universal solution to this type of vulnerability. Author
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Evasion and countermeasures techniques to detect dynamic binary instrumentation frameworks

  • Filho, A. S.
  • Rodríguez, R. J.
  • Feitosa, E. L.
Dynamic Binary Instrumentation (DBI) is a dynamic analysis technique that allows arbitrary code to be executed when a program is running. DBI frameworks have started to be used to analyze malicious applications. As a result, different approaches have merged to detect and avoid them. Commonly referred to as split personality malware or evasive malware are pieces of malicious software that incorporate snippets of code to detect when they are under DBI framework analysis and thus mimic benign behavior. Recent studies have questioned the use of DBI in malware analysis, arguing that it increases the attack surface. In this article, we examine the anti-instrumentation techniques that abuse desktop-based DBI frameworks and existing countermeasures to determine if it is possible to reduce the exploitable attack surface introduced by these DBI frameworks. In particular, we review the related literature to identify (i) the existing set of DBI framework evasion techniques and (ii) the existing set of countermeasures to avoid them. We also analyze and compare the taxonomies introduced in the literature, and propose a new taxonomy that expands and completes the previous taxonomies. Our findings demonstrate that despite advances in DBI framework protections that make them quite suitable for system security purposes, more efforts are needed to reduce the attack surface that they add during application analysis. Only 12 of the 26 evasion techniques covered in this document have countermeasures, threatening the transparency of DBI frameworks. Furthermore, the impact in terms of performance overhead and effectiveness of these countermeasures in real-world situations is unknown. Finally, there are only proofs of concept for 9 of these 26 techniques, which makes it difficult to validate and study how they evade the analysis in order to counter them. We also point out some relevant issues in this context and outline ways of future research directions in the use of DBI frameworks for system security purposes. © 2022 Copyright held by the owner/author(s).
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Towards UAV-based MEC service chain resilience evaluation: a quantitative modeling approach

  • Bai, J.
  • Chang, X.
  • Rodríguez, Ricardo J.
  • Trivedi, K. S.
  • Li, S.
Unmanned aerial vehicle (UAV) and network function virtualization (NFV) facilitate the deployment of multi-access edge computing (MEC). In the UAV-based MEC (UMEC) network, virtualized network function (VNF) can be implemented as a lightweight container running on UMEC host operating system (OS). However, UMEC network is vulnerable to attack, which can result in resource degradation and even UMEC service disruption. Rejuvenation techniques, such as failover technique and live container migration technique, can mitigate the impact of resource degradation but their effectiveness to improve the resilience of UMEC services should be evaluated. This paper presents a quantitative modeling approach based on semi-Markov process to investigate the resilience of a UMEC service chain consisting of any number of VNFs executed in any number of UMEC hosts in terms of availability and reliability. Unlike existing studies, the semi-Markov model constructed in this paper can capture the time-dependent behaviors between VNFs, between host OSes, and between VNFs and host OSes on the condition that the holding times of the recovery and failure events follow any kind of distribution. We perform the sensitivity analysis to identify potential resilience bottlenecks. The results highlight that migration time is the parameter significantly affecting the resilience, which shed the insight on designing the UMEC service chain with high-grade resilience requirements. In addition, we carry out the numerical experiments to reveal that: (i) the type of failure time distribution has a significant effect on the resilience; and (ii) the resilience increases with decreasing number of VNFs, while the availability increases with increasing number of UMEC hosts and the reliability decreases with increasing number of UMEC hosts, which can provide meaningful guidance for the UAV placement optimization in the UMEC network.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Riada: a machine-learning based infrastructure for recognising the emotions of Spotify songs

  • Álvarez, P.
  • García de Quirós, J.
  • Baldassarri, S.
The music emotions can help to improve the personalization of services and contents offered by music streaming providers. Many research works based on the use of machine learning techniques have addressed the problem of recognising the music emotions during the last years. Nevertheless, the results obtained are only applied on small-size music repositories and do not consider what the users feel when they listen to the songs. These issues prevent the existing proposals to be integrated into the personalization mechanisms of the online music providers. In this paper, we present the RIADA infrastructure which is composed by a set of systemsable to annotate emotionally the catalog of songs offered by Spotify based on the users’ perception. RIADA works with the Spotify playlist miner and data services to build emotion recognition models that can solve the open challenges previously mentioned. Machine learning algorithms, music information retrieval techniques, architectures for parallelization of applications and cloud computing have been combined to develop a complex result of engineering able to integrate the music emotions into the Spotify-based applications.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

A learning experience toward the understanding of abstraction-level interactions in parallel applications

  • Valero, Alejandro
  • Gran-Tejero, Rubén
  • Suárez-Gracia, Darío
  • Georgescu, Emanue A.
  • Ezpeleta, Joaquín
  • Álvarez, Pedro
  • Muñoz, Adolfo
  • Ramos, Luis M.
  • Ibáñez, Pablo
In the curriculum of a Computer Engineering program, concepts like parallelism, concurrency, consistency, or atomicity are usually addressed in separate courses due to their thoroughness and extension. Isolating such concepts in courses helps students not only to focus on specific aspects, but also to experience the reality of working with modern computer systems, where those concepts are often detached in different abstraction levels. However, due to such an isolation, it exists a risk of inducing to the students an absence of interactions between these concepts, and, by extension, between the different abstraction levels of a system. This paper proposes a learning experience showcasing the interactions between abstraction levels addressed in laboratory sessions of different courses. The driving example is a parallel ray tracer. In the different courses, students implement and assemble components of this application from the algorithmic level of the tracer to the assembly instructions required to guarantee atomicity. Each lab focuses on a single abstraction level, but shows students the interactions with the rest of the levels. Technical results and student learning outcomes through the analysis of surveys validate the proposed experience and confirm the students learning improvement with a more integrated view of the system.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Module extraction and DLL hijacking detection via single or multiple memory dumps

  • Fernández-Álvarez, Pedro
  • Rodríguez, Ricardo J.
A memory dump contains the current state of a system's physical memory at the time of its acquisition. Among other things, it contains the processes that were running at the time of acquisition. These processes can share certain functionalities provided by shared object files, which are internally represented by modules in Windows. However, each process only maps in its address space the functionalities it needs, and not the entire shared object file. In this way, the current tools for extracting modules from existing processes in a memory dump from a Windows system obtain the partial content of a shared object file instead of the entire file. In this paper we present two tools, dubbed Modex and Intermodex, which are built on top of the Volatility 3 framework. These tools allow a forensic analyst to extract a 64-bit module from one or more Windows memory dumps as completely as possible. To achieve this, they aggregate the contents of the same module loaded by multiple processes that were running in the same memory dump or in different dumps (we called it intradump and interdump, respectively). Additionally, we also show how our developed tools are useful to detect dynamic-link library (DLL) hijacking attacks, a widely used attack on Windows where attackers trick processes into loading a malicious DLL instead of the benign one.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Bringing order to approximate matching: Classification and attacks on similarity digest algorithms

  • Martín-Pérez, M.
  • Rodríguez, R.J.
  • Breitinger, F.
Fuzzy hashing or similarity hashing (a.k.a. bytewise approximate matching) converts digital artifacts into an intermediate representation to allow an efficient (fast) identification of similar objects, e.g., for blacklisting. They gained a lot of popularity over the past decade with new algorithms being developed and released to the digital forensics community. When releasing algorithms (e.g., as part of a scientific article), they are frequently compared with other algorithms to outline the benefits and sometimes also the weaknesses of the proposed approach. However, given the wide variety of algorithms and approaches, it is impossible to provide direct comparisons with all existing algorithms. In this paper, we present the first classification of approximate matching algorithms which allows an easier description and comparisons. Therefore, we first reviewed existing literature to understand the techniques various algorithms use and to familiarize ourselves with the common terminology. Our findings allowed us to develop a categorization relying heavily on the terminology proposed by NIST SP 800-168. In addition to the categorization, this article presents an abstract set of attacks against algorithms and why they are feasible. Lastly, we detail the characteristics needed to build robust algorithms to prevent attacks. We believe that this article helps newcomers, practitioners, and experts alike to better compare algorithms, understand their potential, as well as characteristics and implications they may have on forensic investigations.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Toward Optimal LSTM Neural Networks for Detecting Algorithmically Generated Domain Names

  • Selvi J.
  • Rodríguez Fernández R.J.
  • Soria-Olivas E.
Malware detection is a problem that has become particularly challenging over the last decade. A common strategy for detecting malware is to scan network traffic for malicious connections between infected devices and their command and control (CC) servers. However, malware developers are aware of this detection method and begin to incorporate new strategies to go unnoticed. In particular, they generate domain names instead of using static Internet Protocol addresses or regular domain names pointing to their CC servers. By using a domain generation algorithm, the effectiveness of the blacklisting of domains is reduced, as the large number of domain names that must be blocked greatly increases the size of the blacklist. In this paper, we study different Long Short-Term Memory neural network hyperparameters to find the best network configuration for algorithmically generated domain name detection. In particular, we focus on determining whether the (complex) feature engineering efforts required when using other deep learning techniques, such as Random Forest, can be avoided. In this regard, we have conducted a comparative analysis to study the effect of using different network sizes and configurations on network performance metrics. Our results show an accuracy of 97.62% and an area under the receiver operating characteristic curve of 0.9956 in the test dataset, indicating that it is possible to obtain good classification results despite avoiding the feature engineering process and additional readjustments required in other machine learning techniques.