Séminaire CoaP du 15 Novembre

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons 1 présentation le vendredi 15 novembre prochain. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 4.A101.

Si vous venez participer pour la première fois, n'hésitez pas à contacter les organisateurs pour ne pas être bloqué à l'entrée.

Nicolas Bellec (CEA LIST) - A scalable framework for backward bounded static symbolic execution

Many programs (e.g. malware) hide their behavior by using obfuscations such as opaque predicates. Automatic methods have been developed to detect such obfuscations. In this presentation, we will focus on static symbolic backward bounded execution, a method that enumerates backward bounded paths from a potential opaque predicate and uses symbolic execution to check whether all enumerated paths have the same predicate value. We will show how this method can be generalized using a lattice-based framework and how to scale up such methods to efficiently analyze whole programs.

Séminaire CoaP du 17 juin

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux présentations le lundi 17 juin prochain. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Si vous venez participer pour la première fois, n'hésitez pas à contacter les organisateurs pour ne pas être bloqué à l'entrée.

Arthur Tran Van (Télécom SudParis) - Mealy Verifier: An Automated, Exhaustive, and Explainable Methodology for Analyzing State Machines in Protocol Implementations

Many network protocol specifications are long and lack clarity, which paves the way to implementation errors. Such errors have led to vulnerabilities for secure protocols such as SSH and TLS. Active automata learning, a black-box method, is an efficient method to discover discrepancies between a specification and its implementation. It consists in extracting state machines by interacting with a network stack. It can be (and has been) combined with model checking to analyze the obtained state machines. Model checking is designed for exhibiting a single model violation instead of all model violations and thus leads to a limited understanding of implementation errors. As far as we are aware, there is only one specialized exhaustive method, leveraging DFA (Deterministic Finite Automaton) intersection, available for analyzing the outcomes of active automata learning applied to network protocols. We propose an alternative method, to improve the discovery of new bug and vulnerability patterns and enhance the exhaustiveness of model verification processes. In this presentation, we apply our method to two use cases: OPC UA, for which we present a full workflow from state machine inference to state machine analysis, and SSH, where we focus on the analysis of existing state machines.

Ces travaux seront présentés fin juillet à la conférence ARES.

Shurok Khozam (Télécom SudParis) - DDoS Mitigation while Preserving QoS: A Deep Reinforcement Learning-Based Approach

The deployment of 5G networks has significantly improved connectivity, providing remarkable speed and capacity. These networks rely on Software-Defined Networking (SDN) to enhance control and flexibility. However, this advancement poses critical challenges including expanded attack surface due to network virtualization and the risk of unauthorized access to critical infrastructure. Since traditional cybersecurity methods are inadequate in addressing the dynamic nature of modern cyber attacks, employing artificial intelligence (AI), and deep reinforcement learning (DRL) in particular, was investigated to enhance 5G networks security. This interest arises from the ability of these techniques to dynamically respond and adapt their defense strategies according to encountered situations and real-time threats. Our proposed mitigation system uses a DRL framework, enabling an intelligent agent to dynamically adjust its defense strategies against a range of DDoS attacks, exploiting ICMP, TCP SYN, and UDP, within an SDN environment designed to mirror real-life user behaviors. This approach aims to maintain the network’s performance while concurrently mitigating the impact of the real-time attacks, by providing adaptive and automated countermeasures according to the network’s situation.

Séminaire CoaP du 21 mars

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le jeudi 21 mars prochain. Le séminaire CoaP aura lieu à 14h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Si vous venez participer pour la première fois, n'hésitez pas à contacter les organisateurs pour ne pas être bloqué à l'entrée.

Lorena González-Manzano - Vulnerability detection under poisoning attacks

The complexity of current systems encourages the emergence of vulnerabilities. Detectors are developed in this regard, most of them using Artificial Intelligence (AI) techniques. However, AI is not without its problems, especially those attacks affecting the training set. In this talk a novel vulnerability detector, called VulCoT, is presented, together with their analysis under three different poisoning attacks.

Yanis Sellami - Fault Injection Vulnerability Characterization by Inference of Robust Reachability Constraints

While automated code analysis techniques have succeeded in finding and reporting potential vulnerabilities in binary programs, they tend to report many false positives, which cannot be reliably exploited. This is typical in evaluations of fault injection attacks vulnerabilities as faults can create unexpected program behaviors dependent on complex initial states. As the precise setup of the initial states is hard to achieve, such faults lead code analysis techniques to report vulnerabilities that exist in theory but are infeasible in practice. Vulnerability characterization techniques are thus needed to distinguish such reports from those that come from serious vulnerabilities. Recently, Girol et al. have introduced the concept of robust reachability, a property of program inputs applied to code analysis frameworks to report only vulnerabilities that can be reproduced reliably. This is done by distinguishing inputs that are under the control of the attacker from those that are not, and by reporting only vulnerabilities that do not depend on the value of the uncontrolled inputs. Yet, this remains insufficient for distinguishing severe vulnerabilities from benign ones as robust reachability will be unable to report cases that, e.g., are easy to trigger but may not succeed in a few corner cases. To address this issue, we propose a method that leverages an abduction procedure to generate a robust reachability constraint, that is, a logical constraint on the uncontrolled inputs under which we have the guarantee that the vulnerability will be triggered. We demonstrate the vulnerability characterization capabilities of an implementation of this procedure on a fault injection attack case-study taken from FISSC. We show that our method refines robust reachability and leads to a much better characterization of the reported vulnerabilities. The methods additionally leads to the generation of high-level feedback that is easier to understand and reuse for further analysis.

Slides

Séminaire CoaP du 30 janvier

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le mardi 30 janvier prochain. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Si vous venez participer pour la première fois, n'hésitez pas à contacter les organisateurs pour ne pas être bloqué à l'entrée.

Sara Tucci - Blockchain Consensus Protocols, from Bitcoin to Ethereum 2.0

Bitcoin introduced a fully decentralized, peer-to-peer consensus protocol that enables secure transaction validation in an open network, marking a departure from previous Byzantine Fault Tolerant (BFT) protocols primarily designed for closed networks. An innovative combination of cryptographic and incentive mechanisms ensures the protocol’s robustness over the years. However, it’s important to acknowledge the considerable energy consumption of Bitcoin’s Proof-of-Work mechanism, which remains a significant concern. To address these energy concerns, there have been efforts to transition to more environmentally friendly solutions, such as Proof-of-Stake BFT protocols, like Ethereum 2.0. While these newer proposals hold promise in terms of energy efficiency, they come with complexities and ongoing issues in security and incentive design. In this talk I will present the main features and differences of Proof-Stake-BFT proposals with respect to Bitcoin, to appreciate their maturity and outline open issues and ongoing research challenges.

Adam Oumar ABDEL-RAHMAN - A Privacy-Preserving Infrastructure to Monitor Encrypted DNS Logs

In the realm of cybersecurity, logging system and application activity is a crucial technique to detect and understand cyberattacks by identifying Indicators of Compromise (IoCs). Since these logs can take vast amounts of disk space, it can be tempting to delegate their storage to an external service provider. This requires to encrypt the data, so the service provider does not have access to possibly sensitive information. However, this usually makes it impossible to search for relevant information in the encrypted log. To address this predicament, this paper delves into the realm of modern cryptographic tools to reconcile the dual objectives of protecting log data from prying eyes while enabling controlled processing. We propose a comprehensive framework that contextualizes log data and presents several mechanisms to solve the outsourcing problem, allowing searchable encryption, and we apply our approach to DNS logs. Our contributions include the introduction of two novel schemes, namely symmetric and asymmetric, which facilitate efficient and secure retrieval of intrusion detection-related information from encrypted outsourced storage. Furthermore, we conduct extensive experiments on a test bed to evaluate and compare the effectiveness of the different solutions, providing valuable insights into the practical implementation of our proposed infrastructure for monitoring.

Planches présentées

Séminaire CoaP du 09 novembre

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le jeudi 09 novembre prochain.

Vincent Thouvenot - Privacy attacks against a ResNet50 used for plane classification

In our presentation, we will explain two challenges proposed as part of the CIAD conference: a membership inference attack task and a more original forgetting attack task. We'll explain the first approaches that we used and more importantly, why they failed and the main lessons that we learned from the challenge.

Planches présentées

Matthieu Lemerre - Pas de Crash, Pas d'Exploit: Verification Automatique de Noyaux Embarqués

Le noyau est le composant le plus critique en termes de sûreté et de sécurité de nombreux systèmes informatiques, car un bogue dans celuici peut entraîner un crash ou un exploit système entier. Il est donc souhaitable de garantir qu'un noyau est exempt de ces bogues en utilisant des méthodes formelles, mais le coût élevé et l'expertise requise pour ce faire empêchent une utilisation à large échelle. Nous avons proposé une méthode qui peut vérifier automatiquement à la fois l'absence d'erreurs à l'exécution (c'est-à-dire des crashs) et l'absence d'élévation de privilèges (c'est-à-dire des exploits) dans les noyaux embarqués à partir de leurs exécutables binaires.

Cette méthode a permis de découvrir un système de type permettant de vérifier l'absence de corruption mémoire dans du code C, que nous évoquerons brièvement.

Planches présentées

Séminaire CoaP du 19 septembre

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le mardi 19 septembre prochain. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Quentin Michaud - WebAssembly & Security

WebAssembly (Wasm for short) is a new format of low-level bytecode coming from the Web. It allows to run code sandboxed by default, on a stack-based light virtual machine. It is claiming to bring a lot of dreams to reality : from being the successor of today's containers (by being faster, lighter and more secure), to proposing a single binary format which can be compiled from any programming language and run on any target, without depending on the OS or processor architecture. The promises of Wasm go even beyond technology and address cybersecurity with strong claims regarding the security and protection of Wasm applications. However, articles and publications showing old a new cybersecurity weaknesses inside Wasm may put these claims in doubt. This presentation will give an overview of the Wasm ecosystem, explain the inner workings of Wasm and evaluate the likeliness of its promises as of today and in the future. The promise of Wasm being the successor of containers will be reviewed in more details, both at the container level and at the container orchestrator (Kubernetes) level. The presentation will then propose an assessment of the Wasm claims concerning cybersecurity and take a deeper look at if Wasm can really present itself as an improvement of today binaries' and containers' security.

Bio : Quentin is a last year cybersecurity student at Télécom SudParis and an intern at Thales European research lab ThereSIS, where he is studying bleeding-edge innovations in the cloud ecosystem and their potential uses for cybersecurity. He likes to improve his cybersecurity skills by creating and doing CTFs regularly, and he is consuming and contributing to several open-source projects.

Planches présentées

Frédéric Recoules - What's up in BINSEC? 2022-23 Edition

Software security analyses must often be performed at the executable code level, either because the source code is not available (e.g.: analysis of third-party components, malware or legacy code), or because very low-level attacker models are being considered (hardware or micro-architectural attacks), or because the code must be analyzed after compilation in order to prevent potential compilation bugs or to verify that protections have been properly implemented. Unfortunately, these low-level security analyses are difficult to establish and there are few specialists, hence the need to provide them with the best possible tools via dedicated automated tools.

BINSEC is a formal binary code analysis platform developed at CEA, with a particular focus on security analysis (vulnerabilities, reverse) and the degree of guarantees provided. BINSEC offers original symbolic reasoning engines and multi-architecture support. Recent results have been obtained, for example, in automatic analysis of cryptographic primitives (resistance to covert channel attacks and micro-architectural attacks) or deobfuscation of advanced malware. However, this kind of analysis still suffers from scaling and usability problems.

In this talk, we aim to give an overview of the latest improvements of BINSEC. These advances will be motivated and illustrated through the resolution of various security cases, including recent examples of challenges from the Cyber France Challenge 2022. In particular, we will address problems such as the optimization of a symbolic reasoning engine at the binary level or the symbolic management of self-modifying code. We will also review recent efforts to make the platform more usable (new architectures, simplified initialization, etc.).

Bio : Frédéric Recoules graduated from INSA and Université Toulouse Paul-Sabatier in 2016, then received a PhD in Computer Science from Université Grenoble-Alpes in 2021. His area of expertises spans formal methods, low-level programming, decompilation and reverse engineering. He notably obtained an ICSE distinguished paper award and a 2nd best GDR GPL PhD award (thematic: software engineering, formal methods and programming languages) for his work on formal verification of inline assembly code. He is currently Research Engineer at CEA where he is the main developer and maintainer of the binary-level program analysis platform BINSEC. His research addresses scalability issues in symbolic analysis at binary level, vulnerability analysis and reverse engineering for security.

Séminaire CoaP du 30 mai

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le 30 mai. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Michaël Marcozzi (CEA) - Fine-Grained Coverage-Based Fuzzing

Fuzzing is a popular software testing method that discovers vulnerabilities by massively feeding target applications with automatically generated inputs. Many state-of-art fuzzers use branch coverage as a feedback metric to guide the fuzzing process. The fuzzer retains inputs for further mutation only if branch coverage is increased. However, branch coverage only provides a shallow sampling of program behaviours and hence may discard interesting inputs to mutate. This work aims at taking advantage of the large body of research over defining finer-grained code coverage metrics (such as control-flow, data-flow or mutation coverage) and at evaluating how fuzzing performance is impacted when using these metrics to select interesting inputs for mutation. We propose to make branch coverage-based fuzzers support most fine-grained coverage metrics out of the box (i.e., without changing fuzzer internals). We achieve this by making the test objectives defined by these metrics (such as conditions to activate or mutants to kill) explicit as new branches in the target program. Fuzzing such a modified target is then equivalent to fuzzing the original target, but the fuzzer will also retain inputs covering the additional metrics objectives for mutation. In addition, all the fuzzer mechanisms to penetrate hard-to-cover branches will help covering the additional metrics objectives. We use this approach to evaluate the impact of supporting two fine-grained coverage metrics (multiple condition coverage and weak mutation) over the performance of two state-of-the-art fuzzers (AFL++ and QSYM) with the standard LAVA-M and MAGMA benchmarks. This evaluation suggests that our mechanism for runtime fuzzer guidance, where the fuzzed code is instrumented with additional branches, is effective and could be leveraged to encode guidance from human users or static analysers. Our results also show that the impact of fine-grained metrics over fuzzing performance is hard to predict before fuzzing, and most of the time either neutral or negative. As a consequence, we do not recommend using them to guide fuzzers, except maybe in some possibly favourable circumstances yet to investigate, like for limited parts of the code or to complement classical fuzzing campaigns.

Houda Jmila (TSP) - Analyzing the Vulnerability of Machine Learning-Based IDS to Adversarial Attacks in Cybersecurity

The detection of intrusions is an important aspect of cybersecurity, as it seeks to safeguard computer systems and networks from malicious attacks. While machine learning (ML) techniques have been effective in this field, they face challenges such as the emergence of adversarial attacks that can deceive classifiers. Preventing cybercriminals from exploiting these vulnerabilities is crucial in preventing damage to data and systems. This presentation analyses the vulnerability of both deep learning and shallow classifiers, which are still widely used due to their maturity and ease of implementation, to adversarial attacks in ML-based IDS. Additionally, we explore whether adversarial attacks borrowed from computer vision pose a significant threat to IDS and to what extent realistic adversarial attacks can be generated using these methods.

Séminaire CoaP du 18 Avril

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le 18 avril. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Pierre-Elisée Flory - Comparing Private Set Intersection Various Implementations for Fraud Detection

Banks have to commit answering their customers' privacy concerns while complying to regulation. Sharing information on customer among a Banking consortium is an efficient way to identify fraud at an early stage but requires efficient biometrics matching algorithms to compare two id cards / biometrics template in pictures. Consortium stakeholders may also be competitors and thus need to protect their customer database. Within the Privacy Enhancing Technologies, we have assessed and compared different Secure Multi-Party Computation and in particular Private Set Intersection schemes to mitigate those risks and design a new protocol to allow privacy preserving biometrics templates matching.

Nathanaël Denis - Integrating Usage Control into Distributed Ledger Technology for Internet of Things Privacy

The Internet of Things brings new ways to collect privacy-sensitive data from billions of devices. Well-tailored distributed ledger technologies (DLTs) can provide high transaction processing capacities to IoT devices in a decentralized fashion. However, privacy aspects are often neglected or unsatisfying, with a focus mainly on performance and security. In this paper, we introduce decentralized usage control mechanisms to empower IoT devices to control the data they generate. Usage control defines obligations i.e., actions to be fulfilled to be granted access, and conditions on the system in addition to data dissemination control. The originality of this paper is to consider the usage control system as a component of distributed ledger networks, instead of an external tool. With this integration, both technologies work in synergy, benefiting their privacy, security and performance. We evaluated the performance improvements of integration using the IOTA technology, particularly suitable due to the participation of small devices in the consensus. The results of the tests on a private network show an approximate 90% decrease of the time needed for the UCS to push a transaction and make its access decision in the integrated setting, regardless of the number of nodes in the network.

This contribution is currently under review for publication in a journal.

Séminaire CoaP du 15 février: NVIDIA DOCA hackathon and Adversarial Reachability for Program-level Security Analysis

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le 15 février. Le séminaire CoaP aura lieu à 10h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Romain Ferrari, Louis Cailliot, Julie Sauzedde, Pierre-Elisée Flory - NVIDIA DOCA hackathon

The NVIDIA DOCA hackathon took place on March 21, during NVIDIA 2022 GTC.

The Thales team chose to build a solution upon the DPI acceleration to enable Yara rules, which are used for inspection of files downloaded from the network to identify malware and potential threats. To implement this, Team Thales used a Yara Parser to transform public Yara rules into DPI rules in a Suricata community-based format supported by the DOCA DPI lib. This solution leveraged DOCA DPI functionality to scan the files on the fly as the packets flow through the device.

Soline Ducousso - Adversarial Reachability for Program-level Security Analysis

Many program analysis tools and techniques have been developed to assess program vulnerability. Yet, they are based on the standard concept of reachability and represent an attacker able to craft smart legitimate input, while in practice attackers can be much more powerful, using for instance micro-architectural exploits or fault injection methods. We introduce adversarial reachability , a framework allowing to reason about such advanced attackers and check whether a system is vulnerable or immune to a particular attacker. As equipping the attacker with new capacities significantly increases the state space of the program under analysis, we present a new symbolic exploration algorithm, namely adversarial symbolic execution, injecting faults in a forkless manner to prevent path explosion, together with optimizations dedicated to reduce the number of injections to consider while keeping the same attacker power. Experiments on representative benchmarks from fault injection show that our method significantly reduces the number of adversarial paths to explore, allowing to scale up to 10 faults where prior work timeout for 3 faults. In addition, we analyze the well-tested WooKey's bootloader, and demonstrate the ability of our analysis to find attacks and evaluate countermeasures in real-life security scenarios.

This is joint work with Sébastien Bardin and Marie-Laure Potet.

[Séminaire reporté] Séminaire CoaP du 19 janvier: NVIDIA DOCA hackathon and Adversarial Reachability for Program-level Security Analysis

En raison de l'appel à la grève le 19 janvier, le séminaire est reporté à une date qui doit encore être définie.

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous aurons deux interventions le 19 janvier. Le séminaire CoaP aura lieu à 14h dans le bâtiment IMT/TP/TSP, en salle 3.A213.

Romain Ferrari, Louis Cailliot, Julie Sauzedde, Pierre-Elisée Flory - NVIDIA DOCA hackathon

The NVIDIA DOCA hackathon took place on March 21, during NVIDIA 2022 GTC.

Soline Ducousso - Adversarial Reachability for Program-level Security Analysis

This is joint work with Sébastien Bardin and Marie-Laure Potet.

Séminaire CoaP du 21 novembre

Dans le cadre de notre séminaire « La Cybersécurité sur un plateau » (Cybersecurity on a Plate), nous recevrons le 21 novembre deux intervenants :

Aina Toky Rasoamanana, doctorant à Télécom SudParis, qui présentera ses travaux sur l'inférence de machines à états d'implémentations du protocole TLS.
Mohamad Mansouri, doctorant CIFRE à EURECOM / Thales, qui présentera ses travaux sur l'agrégation sécurisée et tolérante aux pannes pour l'apprentissage fédéré.

Le séminaire CoaP aura lieu à 14h dans le bâtiment IMT/TP/TSP, en salle 3.A405.

Aina Toky Rasoamanana - Towards a Systematic and Automatic Use of State Machine Inference to Uncover Security Flaws and Fingerprint TLS Stacks

TLS is a well-known and thoroughly studied security protocol. In this paper, we focus on a specific class of vulnerabilities affecting TLS implementations, state machine errors. These vulnerabilities are caused by differences in interpreting the standard and correspond to deviations from the specifications, e.g. accepting invalid messages, or accepting valid messages out of sequence. We develop a systematic methodology to infer the state machines of major TLS stacks from stimuli and observations, and to study their evolution across revisions. We use the L* algorithm to compute state machines corresponding to different execution scenarios. We reproduce several known vulnerabilities (denial of service, authentication bypasses), and uncover new ones. We also show that state machine inference is efficient and practical for integration within a continuous integration pipeline, to help find new vulnerabilities or deviations introduced during development.

With our systematic black-box approach, we study over 400 different versions of server and client implementations in various scenarios (protocol version, options). Using the resulting state machines, we propose a robust algorithm to fingerprint TLS stacks. To the best of our knowledge, this is the first application of this approach on such a broad perimeter, in terms of number of TLS stacks, revisions, or execution scenarios studied.

This work has been published at ESORICS 2022.

Mohamad Mansouri - Learning from Failures: Secure and Fault-Tolerant Secure Aggregation for Federated Learning

Federated learning allows multiple parties to collaboratively train a global machine learning (ML) model without sharing their private datasets. To make sure that these local datasets are not leaked, existing works propose to rely on a secure aggregation scheme that allows parties to encrypt their model updates before sending them to the central server that aggregates the encrypted inputs.

In this work, we design and evaluate a new secure and fault-tolerant aggregation scheme for federated learning that is robust against client failures. We first develop a threshold-variant of the secure aggregation scheme proposed by Joye and Libert. Using this new building block together with a dedicated decentralized key management scheme and an input encoding solution, we design a privacy-preserving federated learning protocol that, when executed among n clients, can recover from up to n/3 failures. Our solution is secure against a malicious aggregator who can manipulate messages to learn clients' individual inputs. We show that our solution outperforms the state-of-the-art fault-tolerant secure aggregation schemes in terms of computation cost on the client. For example, with an ML model of 100K parameters, trained with 600 clients, our protocol is 5.5x faster (1.6x faster in case of 180 clients drop).

This work will appear in ACSAC’22.

Grégoire Menguy - Search-Based Local Blackbox Deobfuscation: Understand, Improve and Mitigate (October 4th 2022)

Pour cette première édition, nous recevons Grégoire Menguy, un ancien étudiant de Télécom SudParis, actuellement en thèse au CEA. Son intervention aura lieu à 14h dans le bâtiment IMT/TP/TSP, en salle 3.A405.

Les planches présentées sont disponible via ce lien

Search-Based Local Blackbox Deobfuscation: Understand, Improve and Mitigate

Code obfuscation aims at protecting Intellectual Property and other secrets embedded into software from being retrieved. Recent works leverage advances in artificial intelligence (AI) with the hope of getting blackbox deobfuscators completely immune to standard (whitebox) protection mechanisms. While promising, this new field of AI-based, and more specifically search-based blackbox deobfuscation, is still in its infancy. In this work, we deepen the state of search-based blackbox deobfuscation in three key directions: understand the current state-of-the-art, improve over it and design dedicated protection mechanisms. In particular, we define a novel generic framework for search-based blackbox deobfuscation encompassing prior work and highlighting key components; we are the first to point out that the search space underlying code deobfuscation is too unstable for simulation-based methods (e.g., Monte Carlo Tree Search used in prior work) and advocate the use of robust methods such as S-metaheuristics; we propose the new optimized search-based blackbox deobfuscator Xyntia which significantly outperforms prior work in terms of success rate (especially with small time budget) while being completely immune to the most recent anti-analysis code obfuscation methods; and finally we propose two novel protections against search-based blackbox deobfuscation, allowing to counter Xyntia powerful attacks.

This work has been published at CCS 2021.

Séminaire SCN

Billet dans la catégorie CoaP