arXiv subjects

Murat Kantarcioglu

Publications and source records attributed to Murat Kantarcioglu.

At least 37 records · Page 2Linked to original sources

arXiv2026

MoltGraph: A Longitudinal Temporal Graph Dataset of Moltbook for Coordinated-Agent Detection

Agent-native social platforms such as Moltbook are rapidly emerging, yet they inherit and amplify classical influence and abuse attacks, where coordinated agents strategically comment and upvote to manipulate visibility and propagate narratives across communities. However, rigorous measurement and learning-based monitoring remain constrained by the absence of longitudinal, graph-native datasets for agentic social networks that jointly capture heterogeneous interactions, temporal drift, and visibility signals needed to connect coordination behavior to downstream exposure. We introduce MoltGraph as a realistic longitudinal agentic social-network graph dataset for studying how agents behave, coordinate, and evolve in the wild, enabling reproducible measurement on emerging multi-agent social ecosystems. Using MoltGraph, we provide the first graph-centric characterization of Moltbook as a dynamic network: (i) heavy-tailed connectivity with power-law exponents in the range alpha in [1.86, 2.72], (ii) accelerating hub formation and attention centralization where the top 1% agents account for 29.00% of engagements, (iii) bursty, short-lived coordination episodes, 98.33% last under 24 hours, and (iv) measurable exposure effects across submolts. In matched analyses, posts receiving coordinated engagement exhibit 506.35% higher early interaction rates (within H=5 days) and 242.63% higher downstream exposure in feeds than non-coordinated controls.

cs.SI↗

arXiv2026

Blending Human and LLM Expertise to Detect Hallucinations and Omissions in Mental Health Chatbot Responses

As LLM-powered chatbots are increasingly deployed in mental health services, detecting hallucinations and omissions has become critical for user safety. However, state-of-the-art LLM-as-a-judge methods often fail in high-risk healthcare contexts, where subtle errors can have serious consequences. We show that leading LLM judges achieve only 52% accuracy on mental health counseling data, with some hallucination detection approaches exhibiting near-zero recall. We identify the root cause as LLMs' inability to capture nuanced linguistic and therapeutic patterns recognized by domain experts. To address this, we propose a framework that integrates human expertise with LLMs to extract interpretable, domain-informed features across five analytical dimensions: logical consistency, entity verification, factual accuracy, linguistic uncertainty, and professional appropriateness. Experiments on a public mental health dataset and a new human-annotated dataset show that traditional machine learning models trained on these features achieve 0.717 F1 on our custom dataset and 0.849 F1 on a public benchmark for hallucination detection, with 0.59-0.64 F1 for omission detection across both datasets. Our results demonstrate that combining domain expertise with automated methods yields more reliable and transparent evaluation than black-box LLM judging in high-stakes mental health applications.

cs.CL↗

arXiv2026

Disentangling Prompt Element Level Risk Factors for Hallucinations and Omissions in Mental Health LLM Responses

Mental health concerns are often expressed outside clinical settings, including in high-distress help seeking, where safety-critical guidance may be needed. Consumer health informatics systems increasingly incorporate large language models (LLMs) for mental health question answering, yet many evaluations underrepresent narrative, high-distress inquiries. We introduce UTCO (User, Topic, Context, Tone), a prompt construction framework that represents an inquiry as four controllable elements for systematic stress testing. Using 2,075 UTCO-generated prompts, we evaluated Llama 3.3 and annotated hallucinations (fabricated or incorrect clinical content) and omissions (missing clinically necessary or safety-critical guidance). Hallucinations occurred in 6.5% of responses and omissions in 13.2%, with omissions concentrated in crisis and suicidal ideation prompts. Across regression, element-specific matching, and similarity-matched comparisons, failures were most consistently associated with context and tone, while user-background indicators showed no systematic differences after balancing. These findings support evaluating omissions as a primary safety outcome and moving beyond static benchmark question sets.

cs.CL↗

arXiv2026

SPRINT: Semi-supervised Prototypical Representation for Few-Shot Class-Incremental Tabular Learning

Real-world systems must continuously adapt to novel concepts from limited data without forgetting previously acquired knowledge. While Few-Shot Class-Incremental Learning (FSCIL) is established in computer vision, its application to tabular domains remains largely unexplored. Unlike images, tabular streams (e.g., logs, sensors) offer abundant unlabeled data, a scarcity of expert annotations and negligible storage costs, features ignored by existing vision-based methods that rely on restrictive buffers. We introduce SPRINT, the first FSCIL framework tailored for tabular distributions. SPRINT introduces a mixed episodic training strategy that leverages confidence-based pseudo-labeling to enrich novel class representations and exploits low storage costs to retain base class history. Extensive evaluation across six diverse benchmarks spanning cybersecurity, healthcare, and ecological domains, demonstrates SPRINT's cross-domain robustness. It achieves a state-of-the-art average accuracy of 77.37% (5-shot), outperforming the strongest incremental baseline by 4.45%.

cs.CV↗

arXiv2026

CourtGuard: A Model-Agnostic Framework for Zero-Shot Policy Adaptation in LLM Safety

Current safety mechanisms for Large Language Models (LLMs) rely heavily on static, fine-tuned classifiers that suffer from adaptation rigidity, the inability to enforce new governance rules without expensive retraining. To address this, we introduce CourtGuard, a retrieval-augmented multi-agent framework that reimagines safety evaluation as Evidentiary Debate. By orchestrating an adversarial debate grounded in external policy documents, CourtGuard achieves state-of-the-art performance across 7 safety benchmarks, outperforming dedicated policy-following baselines without fine-tuning. Beyond standard metrics, we highlight two critical capabilities: (1) Zero-Shot Adaptability, where our framework successfully generalized to an out-of-domain Wikipedia Vandalism task (achieving 90\% accuracy) by swapping the reference policy; and (2) Automated Data Curation and Auditing, where we leveraged CourtGuard to curate and audit nine novel datasets of sophisticated adversarial attacks. Our results demonstrate that decoupling safety logic from model weights offers a robust, interpretable, and adaptable path for meeting current and future regulatory requirements in AI governance.

cs.AI↗

arXiv2026

FedDAG: Clustered Federated Learning via Global Data and Gradient Integration for Heterogeneous Environments

Federated Learning (FL) enables a group of clients to collaboratively train a model without sharing individual data, but its performance drops when client data are heterogeneous. Clustered FL tackles this by grouping similar clients. However, existing clustered FL approaches rely solely on either data similarity or gradient similarity; however, this results in an incomplete assessment of client similarities. Prior clustered FL approaches also restrict knowledge and representation sharing to clients within the same cluster. This prevents cluster models from benefiting from the diverse client population across clusters. To address these limitations, FedDAG introduces a clustered FL framework, FedDAG, that employs a weighted, class-wise similarity metric that integrates both data and gradient information, providing a more holistic measure of similarity during clustering. In addition, FedDAG adopts a dual-encoder architecture for cluster models, comprising a primary encoder trained on its own clients' data and a secondary encoder refined using gradients from complementary clusters. This enables cross-cluster feature transfer while preserving cluster-specific specialization. Experiments on diverse benchmarks and data heterogeneity settings show that FedDAG consistently outperforms state-of-the-art clustered FL baselines in accuracy.

cs.LG↗

arXiv2026

Bypassing AI Control Protocols via Agent-as-a-Proxy Attacks

As AI agents automate critical workloads, they remain vulnerable to indirect prompt injection (IPI) attacks. Current defenses rely on monitoring protocols that jointly evaluate an agent's Chain-of-Thought (CoT) and tool-use actions to ensure alignment with user intent. We demonstrate that these monitoring-based defenses can be bypassed via a novel Agent-as-a-Proxy attack, where prompt injection attacks treat the agent as a delivery mechanism, bypassing both agent and monitor simultaneously. While prior work on scalable oversight has focused on whether small monitors can supervise large agents, we show that even frontier-scale monitors are vulnerable. Large-scale monitoring models like Qwen2.5-72B can be bypassed by agents with similar capabilities, such as GPT-4o mini and Llama-3.1-70B. On the AgentDojo benchmark, we achieve a high attack success rate against AlignmentCheck and Extract-and-Evaluate monitors under diverse monitoring LLMs. Our findings suggest current monitoring-based agentic defenses are fundamentally fragile regardless of model scale.

cs.CR↗

arXiv2026

Beyond Refusal: Probing the Limits of Agentic Self-Correction for Semantic Sensitive Information

While defenses for structured PII are mature, Large Language Models (LLMs) pose a new threat: Semantic Sensitive Information (SemSI), where models infer sensitive identity attributes, generate reputation-harmful content, or hallucinate potentially wrong information. The capacity of LLMs to self-regulate these complex, context-dependent sensitive information leaks without destroying utility remains an open scientific question. To address this, we introduce SemSIEdit, an inference-time framework where an agentic "Editor" iteratively critiques and rewrites sensitive spans to preserve narrative flow rather than simply refusing to answer. Our analysis reveals a Privacy-Utility Pareto Frontier, where this agentic rewriting reduces leakage by 34.6% across all three SemSI categories while incurring a marginal utility loss of 9.8%. We also uncover a Scale-Dependent Safety Divergence: large reasoning models (e.g., GPT-5) achieve safety through constructive expansion (adding nuance), whereas capacity-constrained models revert to destructive truncation (deleting text). Finally, we identify a Reasoning Paradox: while inference-time reasoning increases baseline risk by enabling the model to make deeper sensitive inferences, it simultaneously empowers the defense to execute safe rewrites.

cs.AI↗

arXiv2025

LLM-driven Provenance Forensics for Threat Investigation and Detection

We introduce PROVSEEK, an LLM-powered agentic framework for automated provenance-driven forensic analysis and threat intelligence extraction. PROVSEEK employs specialized toolchains to dynamically retrieve relevant context by generating precise, context-aware queries that fuse knowledge from threat reports with evidence from system provenance data. The framework resolves provenance queries, orchestrates multiple role-specific agents, and synthesizes structured, ground-truth verifiable forensic summaries. By combining agent orchestration with Retrieval-Augmented Generation (RAG) and chain-of-thought (CoT) reasoning, data-guided filtration using a behavioral model, PROVSEEK enables adaptive multi-step analysis that iteratively refines hypotheses, verifies supporting evidence, and produces scalable, interpretable forensic explanations of attack behaviors. PROVSEEK is designed for automated threat investigation without task-specific training data, enabling forensic-style investigation even when no prior knowledge of the environment. We conduct a comprehensive evaluation on publicly available DARPA datasets, demonstrating that PROVSEEK outperforms retrieval-based methods for the intelligence extraction task, achieving a 34% improvement in contextual precision/recall; and for threat detection task, PROVSEEK achieves 22%/29% higher precision/recall compared to both a baseline agent approach and State-Of-The-Art (SOTA) Provenance-based Intrusion Detection System (PIDS). In our scalability study, we show that PROVSEEK increases token usage by 1.42x and latency by 1.63x as the database size increases 50x, making it optimal for large-scale deployment. We also conducted an ablation and error analysis study to show how different components of PROVSEEK affect the detection performance.

cs.CR↗

arXiv2025

NOMAD -- Navigating Optimal Model Application to Datastreams

NOMAD (Navigating Optimal Model Application for Datastreams) is an intelligent framework for data enrichment during ingestion that optimizes realtime multiclass classification by dynamically constructing model chains, i.e ,sequences of machine learning models with varying cost-quality tradeoffs, selected via a utilitybased criterion. Inspired by predicate ordering techniques from database query processing, NOMAD leverages cheaper models as initial filters, proceeding to more expensive models only when necessary, while guaranteeing classification quality remains comparable to a designated role model through a formal chain safety mechanism. It employs a dynamic belief update strategy to adapt model selection based on per event predictions and shifting data distributions, and extends to scenarios with dependent models such as earlyexit DNNs and stacking ensembles. Evaluation across multiple datasets demonstrates that NOMAD achieves significant computational savings compared to static and naive approaches while maintaining classification quality comparable to that achieved by the most accurate (and often the most expensive) model.

cs.DB↗

arXiv2025

Learning Joint Embeddings of Function and Process Call Graphs for Malware Detection

Software systems can be represented as graphs, capturing dependencies among functions and processes. An interesting aspect of software systems is that they can be represented as different types of graphs, depending on the extraction goals and priorities. For example, function calls within the software can be captured to create function call graphs, which highlight the relationships between functions and their dependencies. Alternatively, the processes spawned by the software can be modeled to generate process interaction graphs, which focus on runtime behavior and inter-process communication. While these graph representations are related, each captures a distinct perspective of the system, providing complementary insights into its structure and operation. While previous studies have leveraged graph neural networks (GNNs) to analyze software behaviors, most of this work has focused on a single type of graph representation. The joint modeling of both function call graphs and process interaction graphs remains largely underexplored, leaving opportunities for deeper, multi-perspective analysis of software systems. This paper presents a pipeline for constructing and training Function Call Graphs (FCGs) and Process Call Graphs (PCGs) and learning joint embeddings. We demonstrate that joint embeddings outperform a single-graph model. In this paper, we propose GeminiNet, a unified neural network approach that learns joint embeddings from both FCGs and PCGs. We construct a new dataset of 635 Windows executables (318 malicious and 317 benign), extracting FCGs via Ghidra and PCGs via Any.Run sandbox. GeminiNet employs dual graph convolutional branches with an adaptive gating mechanism that balances contributions from static and dynamic views.

cs.LG↗

arXiv2025

Interpreting GNN-based IDS Detections Using Provenance Graph Structural Features

Advanced cyber threats (e.g., Fileless Malware and Advanced Persistent Threat (APT)) have driven the adoption of provenance-based security solutions. These solutions employ Machine Learning (ML) models for behavioral modeling and critical security tasks such as malware and anomaly detection. However, the opacity of ML-based security models limits their broader adoption, as the lack of transparency in their decision-making processes restricts explainability and verifiability. We tailored our solution towards Graph Neural Network (GNN)-based security solutions since recent studies employ GNNs to comprehensively digest system provenance graphs for security-critical tasks. To enhance the explainability of GNN-based security models, we introduce PROVEXPLAINER, a framework offering instance-level security-aware explanations using an interpretable surrogate model. PROVEXPLAINER's interpretable feature space consists of discriminant subgraph patterns and graph structural features, which can be directly mapped to the system provenance problem space, making the explanations human interpretable. We show how PROVEXPLAINER synergizes with current state-of-the-art (SOTA) GNN explainers to deliver domain and instance-specific explanations. We measure the explanation quality using the Fidelity+/Fidelity- metric as used by traditional GNN explanation literature, we incorporate the precision/recall metric, where we consider the accuracy of the explanation against the ground truth, and we designed a human actionability metric based on graph traversal distance. On real-world Fileless and APT datasets, PROVEXPLAINER achieves up to 29%/27%/25%/1.4x higher Fidelity+, precision, recall, and actionability (where higher values are better), and 12% lower Fidelity- (where lower values are better) when compared against SOTA GNN explainers.

cs.CR↗

arXiv2025

Judging with Confidence: Calibrating Autoraters to Preference Distributions

The alignment of large language models (LLMs) with human values increasingly relies on using other LLMs as automated judges, or ``autoraters''. However, their reliability is limited by a foundational issue: they are trained on discrete preference labels, forcing a single ground truth onto tasks that are often subjective, ambiguous, or nuanced. We argue that a reliable autorater must learn to model the full distribution of preferences defined by a target population. In this paper, we propose a general framework for calibrating probabilistic autoraters to any given preference distribution. We formalize the problem and present two learning methods tailored to different data conditions: 1) a direct supervised fine-tuning for dense, probabilistic labels, and 2) a reinforcement learning approach for sparse, binary labels. Our empirical results show that finetuning autoraters with a distribution-matching objective leads to verbalized probability predictions that are better aligned with the target preference distribution, with improved calibration and significantly lower positional bias, all while preserving performance on objective tasks.

cs.CL↗

arXiv2025

Ask ChatGPT: Caveats and Mitigations for Individual Users of AI Chatbots

As ChatGPT and other Large Language Model (LLM)-based AI chatbots become increasingly integrated into individuals' daily lives, important research questions arise. What concerns and risks do these systems pose for individual users? What potential harms might they cause, and how can these be mitigated? In this work, we review recent literature and reports, and conduct a comprehensive investigation into these questions. We begin by explaining how LLM-based AI chatbots work, providing essential background to help readers understand chatbots' inherent limitations. We then identify a range of risks associated with individual use of these chatbots, including hallucinations, intrinsic biases, sycophantic behavior, cognitive decline from overreliance, social isolation, and privacy leakage. Finally, we propose several key mitigation strategies to address these concerns. Our goal is to raise awareness of the potential downsides of AI chatbot use, and to empower users to enhance, rather than diminish, human intelligence, to enrich, rather than compromise, daily life.

cs.CY↗

arXiv2025

PROVCREATOR: Synthesizing Complex Heterogenous Graphs with Node and Edge Attributes

The rise of graph-structured data has driven interest in graph learning and synthetic data generation. While successful in text and image domains, synthetic graph generation remains challenging -- especially for real-world graphs with complex, heterogeneous schemas. Existing research has focused mostly on homogeneous structures with simple attributes, limiting their usefulness and relevance for application domains requiring semantic fidelity. In this research, we introduce ProvCreator, a synthetic graph framework designed for complex heterogeneous graphs with high-dimensional node and edge attributes. ProvCreator formulates graph synthesis as a sequence generation task, enabling the use of transformer-based large language models. It features a versatile graph-to-sequence encoder-decoder that 1. losslessly encodes graph structure and attributes, 2. efficiently compresses large graphs for contextual modeling, and 3. supports end-to-end, learnable graph generation. To validate our research, we evaluate ProvCreator on two challenging domains: system provenance graphs in cybersecurity and knowledge graphs from IntelliGraph Benchmark Dataset. In both cases, ProvCreator captures intricate dependencies between structure and semantics, enabling the generation of realistic and privacy-aware synthetic datasets.

cs.LG↗

arXiv2025

SMOTE-DP: Improving Privacy-Utility Tradeoff with Synthetic Data

Privacy-preserving data publication, including synthetic data sharing, often experiences trade-offs between privacy and utility. Synthetic data is generally more effective than data anonymization in balancing this trade-off, however, not without its own challenges. Synthetic data produced by generative models trained on source data may inadvertently reveal information about outliers. Techniques specifically designed for preserving privacy, such as introducing noise to satisfy differential privacy, often incur unpredictable and significant losses in utility. In this work we show that, with the right mechanism of synthetic data generation, we can achieve strong privacy protection without significant utility loss. Synthetic data generators producing contracting data patterns, such as Synthetic Minority Over-sampling Technique (SMOTE), can enhance a differentially private data generator, leveraging the strengths of both. We prove in theory and through empirical demonstration that this SMOTE-DP technique can produce synthetic data that not only ensures robust privacy protection but maintains utility in downstream learning tasks.

cs.LG↗

arXiv2025

Graph Generative Models Evaluation with Masked Autoencoder

In recent years, numerous graph generative models (GGMs) have been proposed. However, evaluating these models remains a considerable challenge, primarily due to the difficulty in extracting meaningful graph features that accurately represent real-world graphs. The traditional evaluation techniques, which rely on graph statistical properties like node degree distribution, clustering coefficients, or Laplacian spectrum, overlook node features and lack scalability. There are newly proposed deep learning-based methods employing graph random neural networks or contrastive learning to extract graph features, demonstrating superior performance compared to traditional statistical methods, but their experimental results also demonstrate that these methods do not always working well across different metrics. Although there are overlaps among these metrics, they are generally not interchangeable, each evaluating generative models from a different perspective. In this paper, we propose a novel method that leverages graph masked autoencoders to effectively extract graph features for GGM evaluations. We conduct extensive experiments on graphs and empirically demonstrate that our method can be more reliable and effective than previously proposed methods across a number of GGM evaluation metrics, such as "Fréchet Distance (FD)" and "MMD Linear". However, no single method stands out consistently across all metrics and datasets. Therefore, this study also aims to raise awareness of the significance and challenges associated with GGM evaluation techniques, especially in light of recent advances in generative models.

cs.LG↗

arXiv2025

A Review of DeepSeek Models' Key Innovative Techniques

DeepSeek-V3 and DeepSeek-R1 are leading open-source Large Language Models (LLMs) for general-purpose tasks and reasoning, achieving performance comparable to state-of-the-art closed-source models from companies like OpenAI and Anthropic -- while requiring only a fraction of their training costs. Understanding the key innovative techniques behind DeepSeek's success is crucial for advancing LLM research. In this paper, we review the core techniques driving the remarkable effectiveness and efficiency of these models, including refinements to the transformer architecture, innovations such as Multi-Head Latent Attention and Mixture of Experts, Multi-Token Prediction, the co-design of algorithms, frameworks, and hardware, the Group Relative Policy Optimization algorithm, post-training with pure reinforcement learning and iterative training alternating between supervised fine-tuning and reinforcement learning. Additionally, we identify several open questions and highlight potential research opportunities in this rapidly advancing field.

cs.LG↗