arXiv subjects

Christian Wressnegger

Publications and source records attributed to Christian Wressnegger.

18 recordsLinked to original sources

arXiv2026

Anti-Backdoor Coreset Selection via Cumulative Entropy

Recent training-time defenses against neural backdoors isolate a benign subset from poisoned training data, to learn a backdoor-free model from it. In this paper, we formulate this defense strategy as a coreset selection problem, giving rise to so-called "Anti-Backdoor Coreset Selection." Since poisonous samples have (a) lower prediction uncertainty and are (b) less frequent than benign samples, coreset selection naturally focuses more on samples associated with benign functionality than the backdoor functionality. We use the Cumulative Entropy as selection criterion to further facilitate this effect. The metric tracks the learning dynamics of training samples and allowing us to select benign samples with high informativeness for the coreset. Additionally, we unlearn the chosen samples in each epoch to facilitate the separability between benign and poisonous samples. Together, this yields an exceptionally effective training-time defense that constructs a benign coreset to train a backdoor-free model. Unlike prior defenses that compromise natural accuracy and fail against certain attacks, our method mitigates backdooring attacks consistently with a negligible impact on natural performance.

cs.LG↗

arXiv2026

Two Sides of the Same Coin: Learning the Backdoor to Remove the Backdoor

The community has recently developed various training-time defenses to counter neural backdoors introduced through data poisoning. In light of the observation that a model learns poisonous samples responsible for the backdoor easier than benign samples, these approaches either use a fixed threshold of the training loss for splitting or iteratively learn a reference model as an oracle for identifying benign samples. In particular, the latter has proven effective for anti-backdoor learning. Our method, HARVEY, leverages a similar yet crucially different technique: learning an oracle for poisonous rather than benign samples. Learning a backdoored reference model is significantly easier than learning a reference model on benign data. Consequently, we can identify poisonous samples much more accurately than related work identifies benign samples. This crucial difference enables near-perfect backdoor removal as we demonstrate in our evaluation. HARVEY substantially outperforms related approaches across attack types, datasets, and architectures, lowering the attack success rate to the very minimum at a negligible loss in natural accuracy. The figure below shows an overview of our methods working principle.

cs.LG↗

arXiv2026

Rising From the Ashes: How Agentic AI is Unblocking Challenges in Cybersecurity

Security remains a high-cost challenge, with many problems historically deemed inefficient to address or effectively unsolvable. A significant number of these problems stem from labor-intensive tasks that create bottlenecks in defensive approaches. Agentic AI has the potential to alleviate these bottlenecks by directly ingesting and reasoning over natural language or code, thereby expanding the scope of feasible defenses. In this paper, we map open security problems to emergent agentic AI capabilities. To illustrate this potential, we examine 16 case studies, including supply chain analysis, highlighting how agentic AI may benefit defenders.

cs.CR↗

arXiv2025

Chasing Shadows: Pitfalls in LLM Security Research

Large language models (LLMs) are increasingly prevalent in security research. Their unique characteristics, however, introduce challenges that undermine established paradigms of reproducibility, rigor, and evaluation. Prior work has identified common pitfalls in traditional machine learning research, but these studies predate the advent of LLMs. In this paper, we identify nine common pitfalls that have become (more) relevant with the emergence of LLMs and that can compromise the validity of research involving them. These pitfalls span the entire computation process, from data collection, pre-training, and fine-tuning to prompting and evaluation. We assess the prevalence of these pitfalls across all 72 peer-reviewed papers published at leading Security and Software Engineering venues between 2023 and 2024. We find that every paper contains at least one pitfall, and each pitfall appears in multiple papers. Yet only 15.7% of the present pitfalls were explicitly discussed, suggesting that the majority remain unrecognized. To understand their practical impact, we conduct four empirical case studies showing how individual pitfalls can mislead evaluation, inflate performance, or impair reproducibility. Based on our findings, we offer actionable guidelines to support the community in future work.

cs.CR↗

arXiv2025

S2AP: Score-space Sharpness Minimization for Adversarial Pruning

Adversarial pruning methods have emerged as a powerful tool for compressing neural networks while preserving robustness against adversarial attacks. These methods typically follow a three-step pipeline: (i) pretrain a robust model, (ii) select a binary mask for weight pruning, and (iii) finetune the pruned model. To select the binary mask, these methods minimize a robust loss by assigning an importance score to each weight, and then keep the weights with the highest scores. However, this score-space optimization can lead to sharp local minima in the robust loss landscape and, in turn, to an unstable mask selection, reducing the robustness of adversarial pruning methods. To overcome this issue, we propose a novel plug-in method for adversarial pruning, termed Score-space Sharpness-aware Adversarial Pruning (S2AP). Through our method, we introduce the concept of score-space sharpness minimization, which operates during the mask search by perturbing importance scores and minimizing the corresponding robust loss. Extensive experiments across various datasets, models, and sparsity levels demonstrate that S2AP effectively minimizes sharpness in score space, stabilizing the mask selection, and ultimately improving the robustness of adversarial pruning methods.

cs.CV↗

arXiv2025

Controlling Latent Diffusion Using Latent CLIP

Instead of performing text-conditioned denoising in the image domain, latent diffusion models (LDMs) operate in latent space of a variational autoencoder (VAE), enabling more efficient processing at reduced computational costs. However, while the diffusion process has moved to the latent space, the contrastive language-image pre-training (CLIP) models, as used in many image processing tasks, still operate in pixel space. Doing so requires costly VAE-decoding of latent images before they can be processed. In this paper, we introduce Latent-CLIP, a CLIP model that operates directly in the latent space. We train Latent-CLIP on 2.7B pairs of latent images and descriptive texts, and show that it matches zero-shot classification performance of similarly sized CLIP models on both the ImageNet benchmark and a LDM-generated version of it, demonstrating its effectiveness in assessing both real and generated content. Furthermore, we construct Latent-CLIP rewards for reward-based noise optimization (ReNO) and show that they match the performance of their CLIP counterparts on GenEval and T2I-CompBench while cutting the cost of the total pipeline by 21%. Finally, we use Latent-CLIP to guide generation away from harmful content, achieving strong performance on the inappropriate image prompts (I2P) benchmark and a custom evaluation, without ever requiring the costly step of decoding intermediate images.

cs.CV↗

arXiv2024

Holistic Adversarially Robust Pruning

Neural networks can be drastically shrunk in size by removing redundant parameters. While crucial for the deployment on resource-constraint hardware, oftentimes, compression comes with a severe drop in accuracy and lack of adversarial robustness. Despite recent advances, counteracting both aspects has only succeeded for moderate compression rates so far. We propose a novel method, HARP, that copes with aggressive pruning significantly better than prior work. For this, we consider the network holistically. We learn a global compression strategy that optimizes how many parameters (compression rate) and which parameters (scoring connections) to prune specific to each layer individually. Our method fine-tunes an existing model with dynamic regularization, that follows a step-wise incremental function balancing the different objectives. It starts by favoring robustness before shifting focus on reaching the target compression rate and only then handles the objectives equally. The learned compression strategies allow us to maintain the pre-trained model natural accuracy and its adversarial robustness for a reduction by 99% of the network original size. Moreover, we observe a crucial influence of non-uniform compression across layers.

cs.LG↗

arXiv2024

Generalized Adversarial Code-Suggestions: Exploiting Contexts of LLM-based Code-Completion

While convenient, relying on LLM-powered code assistants in day-to-day work gives rise to severe attacks. For instance, the assistant might introduce subtle flaws and suggest vulnerable code to the user. These adversarial code-suggestions can be introduced via data poisoning and, thus, unknowingly by the model creators. In this paper, we provide a generalized formulation of such attacks, spawning and extending related work in this domain. This formulation is defined over two components: First, a trigger pattern occurring in the prompts of a specific user group, and, second, a learnable map in embedding space from the prompt to an adversarial bait. The latter gives rise to novel and more flexible targeted attack-strategies, allowing the adversary to choose the most suitable trigger pattern for a specific user-group arbitrarily, without restrictions on the pattern's tokens. Our directional-map attacks and prompt-indexing attacks increase the stealthiness decisively. We extensively evaluate the effectiveness of these attacks and carefully investigate defensive mechanisms to explore the limits of generalized adversarial code-suggestions. We find that most defenses unfortunately offer little protection only.

cs.CR↗

arXiv2023

Machine Unlearning of Features and Labels

Removing information from a machine learning model is a non-trivial task that requires to partially revert the training process. This task is unavoidable when sensitive data, such as credit card numbers or passwords, accidentally enter the model and need to be removed afterwards. Recently, different concepts for machine unlearning have been proposed to address this problem. While these approaches are effective in removing individual data points, they do not scale to scenarios where larger groups of features and labels need to be reverted. In this paper, we propose the first method for unlearning features and labels. Our approach builds on the concept of influence functions and realizes unlearning through closed-form updates of model parameters. It enables to adapt the influence of training data on a learning model retrospectively, thereby correcting data leaks and privacy issues. For learning models with strongly convex loss functions, our method provides certified unlearning with theoretical guarantees. For models with non-convex losses, we empirically show that unlearning features and labels is effective and significantly faster than other strategies.

cs.LG↗

arXiv2022

Backdooring Explainable Machine Learning

Explainable machine learning holds great potential for analyzing and understanding learning-based systems. These methods can, however, be manipulated to present unfaithful explanations, giving rise to powerful and stealthy adversaries. In this paper, we demonstrate blinding attacks that can fully disguise an ongoing attack against the machine learning model. Similar to neural backdoors, we modify the model's prediction upon trigger presence but simultaneously also fool the provided explanation. This enables an adversary to hide the presence of the trigger or point the explanation to entirely different portions of the input, throwing a red herring. We analyze different manifestations of such attacks for different explanation types in the image domain, before we resume to conduct a red-herring attack against malware classification.

cs.CR↗

arXiv2021

Dos and Don'ts of Machine Learning in Computer Security

With the growing processing power of computing systems and the increasing availability of massive datasets, machine learning algorithms have led to major breakthroughs in many different areas. This development has influenced computer security, spawning a series of work on learning-based security systems, such as for malware detection, vulnerability discovery, and binary code analysis. Despite great potential, machine learning in security is prone to subtle pitfalls that undermine its performance and render learning-based systems potentially unsuitable for security tasks and practical deployment. In this paper, we look at this problem with critical eyes. First, we identify common pitfalls in the design, implementation, and evaluation of learning-based security systems. We conduct a study of 30 papers from top-tier security conferences within the past 10 years, confirming that these pitfalls are widespread in the current security literature. In an empirical analysis, we further demonstrate how individual pitfalls can lead to unrealistic performance and interpretations, obstructing the understanding of the security problem at hand. As a remedy, we propose actionable recommendations to support researchers in avoiding or mitigating the pitfalls where possible. Furthermore, we identify open problems when applying machine learning in security and provide directions for further research.

cs.CR↗

arXiv2021

LaserShark: Establishing Fast, Bidirectional Communication into Air-Gapped Systems

Physical isolation, so called air-gapping, is an effective method for protecting security-critical computers and networks. While it might be possible to introduce malicious code through the supply chain, insider attacks, or social engineering, communicating with the outside world is prevented. Different approaches to breach this essential line of defense have been developed based on electromagnetic, acoustic, and optical communication channels. However, all of these approaches are limited in either data rate or distance, and frequently offer only exfiltration of data. We present a novel approach to infiltrate data to and exfiltrate data from air-gapped systems without any additional hardware on-site. By aiming lasers at already built-in LEDs and recording their response, we are the first to enable a long-distance (25m), bidirectional, and fast (18.2kbps in & 100kbps out) covert communication channel. The approach can be used against any office device that operates LEDs at the CPU's GPIO interface.

cs.CR↗

arXiv2020

Evaluating Explanation Methods for Deep Learning in Security

Deep learning is increasingly used as a building block of security systems. Unfortunately, neural networks are hard to interpret and typically opaque to the practitioner. The machine learning community has started to address this problem by developing methods for explaining the predictions of neural networks. While several of these approaches have been successfully applied in the area of computer vision, their application in security has received little attention so far. It is an open question which explanation methods are appropriate for computer security and what requirements they need to satisfy. In this paper, we introduce criteria for comparing and evaluating explanation methods in the context of computer security. These cover general properties, such as the accuracy of explanations, as well as security-focused aspects, such as the completeness, efficiency, and robustness. Based on our criteria, we investigate six popular explanation methods and assess their utility in security systems for malware detection and vulnerability discovery. We observe significant differences between the methods and build on these to derive general recommendations for selecting and applying explanation methods in computer security.

cs.LG↗

arXiv2020

Aim Low, Shoot High: Evading Aimbot Detectors by Mimicking User Behavior

Current schemes to detect cheating in online games often build on the assumption that the applied cheat takes actions that are drastically different from normal behavior. For instance, an Aimbot for a first-person shooter is used by an amateur player to increase his/her capabilities many times over. Attempts to evade detection would require to reduce the intended effect such that the advantage is presumably lowered into insignificance. We argue that this is not necessarily the case and demonstrate how a professional player is able to make use of an adaptive Aimbot that mimics user behavior to gradually increase performance and thus evades state-of-the-art detection mechanisms. We show this in a quantitative and qualitative evaluation with two professional "Counter-Strike: Global Offensive" players, two open-source Anti-Cheat systems, and the commercially established combination of VAC, VACnet, and Overwatch.

cs.CR↗

arXiv2019

Political Elections Under (Social) Fire? Analysis and Detection of Propaganda on Twitter

For many, social networks have become the primary source of news, although the correctness of the provided information and its trustworthiness are often unclear. The investigations of the 2016 US presidential elections have brought the existence of external campaigns to light aiming at affecting the general political public opinion. In this paper, we investigate whether a similar influence on political elections can be observed in Europe as well. To this end, we use the past German federal election as an indicator and inspect the propaganda on Twitter, based on data from a period of 268 days. We find that 79 trolls from the US campaign have also acted upon the German federal election spreading right-wing views. Moreover, we develop a detector for finding automated behavior that enables us to identify 2,414 previously unknown bots.

cs.CR↗

arXiv2018

Poisoning Behavioral Malware Clustering

Clustering algorithms have become a popular tool in computer security to analyze the behavior of malware variants, identify novel malware families, and generate signatures for antivirus systems. However, the suitability of clustering algorithms for security-sensitive settings has been recently questioned by showing that they can be significantly compromised if an attacker can exercise some control over the input data. In this paper, we revisit this problem by focusing on behavioral malware clustering approaches, and investigate whether and to what extent an attacker may be able to subvert these approaches through a careful injection of samples with poisoning behavior. To this end, we present a case study on Malheur, an open-source tool for behavioral malware clustering. Our experiments not only demonstrate that this tool is vulnerable to poisoning attacks, but also that it can be significantly compromised even if the attacker can only inject a very small percentage of attacks into the input data. As a remedy, we discuss possible countermeasures and highlight the need for more secure clustering algorithms.

cs.LG↗

arXiv2018

Web-based Cryptojacking in the Wild

With the introduction of memory-bound cryptocurrencies, such as Monero, the implementation of mining code in browser-based JavaScript has become a worthwhile alternative to dedicated mining rigs. Based on this technology, a new form of parasitic computing, widely called cryptojacking or drive-by mining, has gained momentum in the web. A cryptojacking site abuses the computing resources of its visitors to covertly mine for cryptocurrencies. In this paper, we systematically explore this phenomenon. For this, we propose a 3-phase analysis approach, which enables us to identify mining scripts and conduct a large-scale study on the prevalence of cryptojacking in the Alexa 1 million websites. We find that cryptojacking is common, with currently 1 out of 500 sites hosting a mining script. Moreover, we perform several secondary analyses to gain insight into the cryptojacking landscape, including a measurement of code characteristics, an estimate of expected mining revenue, and an evaluation of current blacklist-based countermeasures.

cs.CR↗

arXiv2016

From Malware Signatures to Anti-Virus Assisted Attacks

Although anti-virus software has significantly evolved over the last decade, classic signature matching based on byte patterns is still a prevalent concept for identifying security threats. Anti-virus signatures are a simple and fast detection mechanism that can complement more sophisticated analysis strategies. However, if signatures are not designed with care, they can turn from a defensive mechanism into an instrument of attack. In this paper, we present a novel method for automatically deriving signatures from anti-virus software and demonstrate how the extracted signatures can be used to attack sensible data with the aid of the virus scanner itself. We study the practicability of our approach using four commercial products and exemplarily discuss a novel attack vector made possible by insufficiently designed signatures. Our research indicates that there is an urgent need to improve pattern-based signatures if used in anti-virus software and to pursue alternative detection approaches in such products.

cs.CR↗