Helmholtz Center for Information Security
CISPA – Helmholtz-Zentrum für InformationssicherheitNot a member yet
3406 research outputs found
Sort by
GAN-Leaks: A Taxonomy of Membership Inference Attacks against Generative Models
Get PDFDeep learning has achieved overwhelming success, spanning from discriminative models to generative models. In particular, deep generative models have facilitated a new level of performance in a myriad of areas, ranging from media manipulation to sanitized dataset generation. Despite the great success, the potential risks of privacy breach caused by generative models have not been analyzed systematically. In this paper, we focus on membership inference attack against deep generative models that reveals information about the training data used for victim models. Specifically, we present the first taxonomy of membership inference attacks, encompassing not only existing attacks but also our novel ones. In addition, we propose the first generic attack model that can be instantiated in a large range of settings and is applicable to various kinds of deep generative models. Moreover, we provide a theoretically grounded attack calibration technique, which consistently boosts the attack performance in all cases, across different attack settings, data modalities, and training configurations. We complement the systematic analysis of attack performance by a comprehensive experimental study, that investigates the effectiveness of various attacks w.r.t. model type and training configurations, over three diverse application scenarios (i.e., images, medical data, and location data)
Monitoring Cyber-Physical Systems: From Design to Integration
Get PDFCyber-physical systems are inherently safety-critical. The deployment of a runtime monitor significantly increases confidence in their safety. The effectiveness of the monitor can be maximized by considering it an integral component during its development. Thus, in this paper, I given an overview over recent work regarding a development process for runtime monitors alongside a cyber-physical system. This process includes the transformation of desirable safety properties into the formal specification language RTLola. A compiler then generates an executable artifact for monitoring the specification. This artifact can then be integrated into the system
Limiting the impact of unreliable randomness in deployed security protocols
Get PDFMany cryptographic mechanisms depend upon the availability of securely generated random numbers. In practice,the sources of random numbers can be unreliable for many reasons, including bugs, compromise or subversion of standards. While there exist ways to significantly reduce the impact of unreliable randomness, these typically do not work well with practical constraints, such as long-term keys stored in hardware security modules. In practice, even modern protocols like TLS 1.3 lack such mechanisms and are therefore highly vulnerable to unreliable randomness.We propose a wrapper construction that reduces the impact of untrusted randomness, and which is is compatible with, and effective in, existing deployments of protocols such as TLS. We provide a security analysis of the construction and elaborate on design choices and practical interpretations. Our findings show that it is possible to effectively harden deployed protocols against unreliable randomness
Prediction Poisoning: Utility-Constrained Defenses Against Model Stealing Attacks
No full textHigh-performance Deep Neural Networks (DNNs) are increasingly deployed in many real-world applications e.g., cloud prediction APIs. Recent advances in model functionality stealing attacks via black-box access (i.e., inputs in, predictions out) threaten the business model of such applications, which require a lot of time, money, and effort to develop. Existing defenses take a passive role against stealing attacks, such as by truncating predicted information. We find such passive defenses ineffective against DNN stealing attacks. In this paper, we propose the first defense which actively perturbs predictions targeted at poisoning the training objective of the attacker. We find our defense effective across a wide range of challenging datasets and DNN model stealing attacks, and additionally outperforms existing defenses. Our defense is the first that can withstand highly accurate model stealing attacks for tens of thousands of queries, amplifying the attacker's error rate up to a factor of 85x with minimal impact on the utility for benign users
Haar Wavelet based Block Autoregressive Flows for Trajectories
Get PDFPrediction of trajectories such as that of pedestrians is crucial to the performance of autonomous agents. While previous works have leveraged conditional generative models like GANs and VAEs for learning the likely future trajectories, accurately modeling the dependency structure of these multimodal distributions, particularly over long time horizons remains challenging. Normalizing flow based generative models can model complex distributions admitting exact inference. These include variants with split coupling invertible transformations that are easier to parallelize compared to their autoregressive counterparts. To this end, we introduce a novel Haar wavelet based block autoregressive model leveraging split couplings, conditioned on coarse trajectories obtained from Haar wavelet based transformations at different levels of granularity. This yields an exact inference method that models trajectories at different spatio-temporal resolutions in a hierarchical manner. We illustrate the advantages of our approach for generating diverse and accurate trajectories on two real-world datasets - Stanford Drone and Intersection Drone
Studying JavaScript Security Through Static Analysis
Get PDFAs the Internet keeps on growing, so does the interest of malicious actors. While the Internet has become widespread and popular to interconnect billions of people, this interconnectivity also simplifies the spread of malicious software. Specifically, JavaScript has become a popular attack vector, as it enables to stealthily exploit bugs and further vulnerabilities to compromise the security and privacy of Internet users. In this thesis, we approach these issues by proposing several systems to statically analyze real-world JavaScript code at scale.
First, we focus on the detection of malicious JavaScript samples. To this end, we propose two learning-based pipelines, which leverage syntactic, control and data-flow based features to distinguish benign from malicious inputs. Subsequently, we evaluate the robustness of such static malicious JavaScript detectors in an adversarial setting. For this purpose, we introduce a generic camouflage attack, which consists in rewriting malicious samples to reproduce existing benign syntactic structures. Finally, we consider vulnerable browser extensions. In particular, we abstract an extension source code at a semantic level, including control, data, and message flows, and pointer analysis, to detect suspicious data flows from and toward an extension privileged context. Overall, we report on 184 Chrome extensions that attackers could exploit to, e.g., execute arbitrary code in a victim's browser
IReEn: Iterative Reverse-Engineering of Black-Box Functions via Neural Program Synthesis
Get PDFIn this work, we investigate the problem of revealing the functionality of a black-box agent. Notably, we are interested in the interpretable and formal description of the behavior of such an agent. Ideally, this description would take the form of a program written in a high-level language. This task is also known as reverse engineering and plays a pivotal role in software engineering, computer security, but also most recently in interpretability. In contrast to prior work, we do not rely on privileged information on the black box, but rather investigate the problem under a weaker assumption of having only access to inputs and outputs of the program. We approach this problem by iteratively refining a candidate set using a generative neural program synthesis approach until we arrive at a functionally equivalent program. We assess the performance of our approach on the Karel dataset. Our results show that the proposed approach outperforms the state-of-the-art on this challenge by finding a functional equivalent program in 78% of cases -- even exceeding prior work that had privileged information on the black-box