Lukas Pirch

Lukas Pirch, Micha Horlboge, Patrick Großmann et al.

Autonomous agents based on large language models (LLMs) are rapidly emerging as a general-purpose technology, with recent systems such as OpenClaw extending their capabilities through broad tool use, third-party skills, and deeper integration into user environments. At the same time, these agentic systems introduce substantial security risks by combining unconstrained capabilities with access to sensitive user data. In this work, we investigate the security of LLM-based agents through the lens of operating systems. We argue that both face strikingly similar challenges in isolating resources, separating privileges, and mediating communication. Guided by this perspective, we survey the current landscape of open-source agents, derive a unified agent architecture, and systematically analyze potential attack vectors. To validate this analysis, we conduct a case study evaluating four widely used OpenClaw-like agents. Even under modest attacker capabilities, we find that several protection mechanisms fail in practice and that secure operation requires detailed system knowledge and careful configuration. However, we also observe that while some agentic capabilities remain insecure by design, many vulnerabilities can be mitigated using well-established techniques from operating system security. We conclude with a set of recommendations for the secure design of agentic systems.

Dilyara Bareeva, Marina M. -C. Höhne, Alexander Warnecke et al.

Feature Visualization (FV) is a widely used technique for interpreting the concepts learned by Deep Neural Networks (DNNs), which synthesizes input patterns that maximally activate a given feature. Despite its popularity, the trustworthiness of FV explanations has received limited attention. In this paper, we introduce a novel method, Gradient Slingshots, that enables manipulation of FV without modifying the model architecture or significantly degrading its performance. By shaping new trajectories in the off-distribution regions of the activation landscape of a feature, we coerce the optimization process to converge in a predefined visualization. We evaluate our approach on several DNN architectures, demonstrating its ability to replace faithfuls FV with arbitrary targets. These results expose a critical vulnerability: auditors relying solely on FV may accept entirely fabricated explanations. To mitigate this risk, we propose a straightforward defense and quantitatively demonstrate its effectiveness.

Felix Weissberg, Lukas Pirch, Erik Imgrund et al.

Large language models (LLMs) excel in many tasks of software engineering, yet progress in leveraging them for vulnerability discovery has stalled in recent years. To understand this phenomenon, we investigate LLMs through the lens of classic code metrics. Surprisingly, we find that a classifier trained solely on these metrics performs on par with state-of-the-art LLMs for vulnerability discovery. A root-cause analysis reveals a strong correlation and a causal effect between LLMs and code metrics: When the value of a metric is changed, LLM predictions tend to shift by a corresponding magnitude. This dependency suggests that LLMs operate at a similarly shallow level as code metrics, limiting their ability to grasp complex patterns and fully realize their potential in vulnerability discovery. Based on these findings, we derive recommendations on how research should more effectively address this challenge.

Alexander Warnecke, Lukas Pirch, Christian Wressnegger et al.

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.

Erwin Quiring, Lukas Pirch, Michael Reimsbach et al.

Machine learning-based systems for malware detection operate in a hostile environment. Consequently, adversaries will also target the learning system and use evasion attacks to bypass the detection of malware. In this paper, we outline our learning-based system PEberus that got the first place in the defender challenge of the Microsoft Evasion Competition, resisting a variety of attacks from independent attackers. Our system combines multiple, diverse defenses: we address the semantic gap, use various classification models, and apply a stateful defense. This competition gives us the unique opportunity to examine evasion attacks under a realistic scenario. It also highlights that existing machine learning methods can be hardened against attacks by thoroughly analyzing the attack surface and implementing concepts from adversarial learning. Our defense can serve as an additional baseline in the future to strengthen the research on secure learning.