arXiv subjects

Nidhi Rastogi

Publications and source records attributed to Nidhi Rastogi.

At least 19 recordsLinked to original sources

arXiv2026

Minerva: Reinforcement Learning with Verifiable Rewards for Cyber Threat Intelligence LLMs

Cyber threat intelligence (CTI) analysts routinely convert noisy, unstructured security artifacts into standardized, automation-ready representations. Although large language models (LLMs) show promise for this task, existing approaches remain brittle when producing structured CTI outputs and have largely relied on supervised fine-tuning (SFT). In contrast, CTI standards and community-maintained resources define canonical identifiers and schemas that enable deterministic verification of model outputs. We leverage this structure to study reinforcement learning with verifiable rewards (RLVR) for CTI tasks. We introduce Minerva, a unified dataset and training pipeline spanning multiple CTI subtasks, each paired with task-specific verifiers that score structured outputs and identifier predictions. To address reward sparsity during rollout, we propose MinervaRL, a lightweight self-training mechanism that generates additional verified trajectories and distills them back into the model. Averaged across four backbones and 12 CTI benchmarks, MinervaRL improves the mean score by 15.8 percentage points over the corresponding base models and by 4.3 points over GRPO.

cs.LG↗

arXiv2026

H-Sets: Hessian-Guided Discovery of Set-Level Feature Interactions in Image Classifiers

Feature attribution methods explain the predictions of deep neural networks by assigning importance scores to individual input features. However, most existing methods focus solely on marginal effects, overlooking feature interactions, where groups of features jointly influence model output. Such interactions are especially important in image classification tasks, where semantic meaning often arises from pixel interdependencies rather than isolated features. Existing interaction-based methods for images are either coarse (e.g., superpixel-only) or, fail to satisfy core interpretability axioms. In this work, we introduce H-Sets, a novel two-stage framework for discovering and attributing higher-order feature interactions in image classifiers. First, we detect locally interacting pairs via input Hessians and recursively merge them into semantically coherent sets; segmentation from Segment Anything (SAM) is used as a spatial grouping prior but can be replaced by other segmentations. Second, we attribute each set with IDG-Vis, a set-level extension of Integrated Directional Gradients that integrates directional gradients along pixel-space paths and aggregates them with Harsanyi dividends. While Hessians introduce additional compute at the detection stage, this targeted cost consistently yields saliency maps that are sparser and more faithful. Evaluations across VGG, ResNet, DenseNet and MobileNet models on ImageNet and CUB datasets show that H-Sets generate more interpretable and faithful saliency maps compared to existing methods.

cs.CV↗

arXiv2026

SPHINX: A Synthetic Environment for Visual Perception and Reasoning

We present Sphinx, a synthetic environment for visual perception and reasoning that targets core cognitive primitives. Sphinx procedurally generates puzzles using motifs, tiles, charts, icons, and geometric primitives, each paired with verifiable ground-truth solutions, enabling both precise evaluation and large-scale dataset construction. The benchmark covers 25 task types spanning symmetry detection, geometric transformations, spatial reasoning, chart interpretation, and sequence prediction. Evaluating recent large vision-language models (LVLMs) shows that even state-of-the-art GPT-5 attains only 51.1% accuracy, well below human performance. Finally, we demonstrate that reinforcement learning with verifiable rewards (RLVR) substantially improves model accuracy on these tasks and yields gains on external visual reasoning benchmarks, highlighting its promise for advancing multimodal reasoning.

cs.CV↗

arXiv2026

Training for Trustworthy Saliency Maps: Adversarial Training Meets Feature-Map Smoothing

Gradient-based saliency methods such as Vanilla Gradient (VG) and Integrated Gradients (IG) are widely used to explain image classifiers, yet the resulting maps are often noisy and unstable, limiting their usefulness in high-stakes settings. Most prior work improves explanations by modifying the attribution algorithm, leaving open how the training procedure shapes explanation quality. We take a training-centered view and first provide a curvature-based analysis linking attribution stability to how smoothly the input-gradient field varies locally. Guided by this connection, we study adversarial training and identify a consistent trade-off: it yields sparser and more input-stable saliency maps, but can degrade output-side stability, causing explanations to change even when predictions remain unchanged and logits vary only slightly. To mitigate this, we propose augmenting adversarial training with a lightweight feature-map smoothing block that applies a differentiable Gaussian filter in an intermediate layer. Across FMNIST, CIFAR-10, and ImageNette, our method preserves the sparsity benefits of adversarial training while improving both input-side stability and output-side stability. A human study with 65 participants further shows that smoothed adversarial saliency maps are perceived as more sufficient and trustworthy. Overall, our results demonstrate that explanation quality is critically shaped by training, and that simple smoothing with robust training provides a practical path toward saliency maps that are both sparse and stable.

cs.CV↗

arXiv2026

AthenaBench: A Dynamic Benchmark for Evaluating LLMs in Cyber Threat Intelligence

Large Language Models (LLMs) have demonstrated strong capabilities in natural language reasoning, yet their application to Cyber Threat Intelligence (CTI) remains limited. CTI analysis involves distilling large volumes of unstructured reports into actionable knowledge, a process where LLMs could substantially reduce analyst workload. CTIBench introduced a comprehensive benchmark for evaluating LLMs across multiple CTI tasks. In this work, we extend CTIBench by developing AthenaBench, an enhanced benchmark that includes an improved dataset creation pipeline, duplicate removal, refined evaluation metrics, and a new task focused on risk mitigation strategies. We evaluate twelve LLMs, including state-of-the-art proprietary models such as GPT-5 and Gemini-2.5 Pro, alongside seven open-source models from the LLaMA and Qwen families. While proprietary LLMs achieve stronger results overall, their performance remains subpar on reasoning-intensive tasks, such as threat actor attribution and risk mitigation, with open-source models trailing even further behind. These findings highlight fundamental limitations in the reasoning capabilities of current LLMs and underscore the need for models explicitly tailored to CTI workflows and automation.

cs.CR↗

arXiv2025

PROVEX: Enhancing SOC Analyst Trust with Explainable Provenance-Based IDS

Modern intrusion detection systems (IDS) leverage graph neural networks (GNNs) to detect malicious activity in system provenance data, but their decisions often remain a black box to analysts. This paper presents a comprehensive XAI framework designed to bridge the trust gap in Security Operations Centers (SOCs) by making graph-based detection transparent. We implement this framework on top of KAIROS, a state-of-the-art temporal graph-based IDS, though our design is applicable to any temporal graph-based detector with minimal adaptation. The complete codebase is available at https://github.com/devang1304/provex.git. We augment the detection pipeline with post-hoc explanations that highlight why an alert was triggered, identifying key causal subgraphs and events. We adapt three GNN explanation methods - GraphMask, GNNExplainer, and a variational temporal GNN explainer (VA-TGExplainer) - to the temporal provenance context. These tools output human-interpretable representations of anomalous behavior, including important edges and uncertainty estimates. Our contributions focus on the practical integration of these explainers, addressing challenges in memory management and reproducibility. We demonstrate our framework on the DARPA CADETS Engagement 3 dataset and show that it produces concise window-level explanations for detected attacks. Our evaluation reveals that the explainers preserve the TGNN's decisions with high fidelity, surfacing critical edges such as malicious file interactions and anomalous netflows. The average explanation overhead is 3-5 seconds per event. By providing insight into the model's reasoning, our framework aims to improve analyst trust and triage speed.

cs.CR↗

arXiv2025

Limits of Generalization in RLVR: Two Case Studies in Mathematical Reasoning

Mathematical reasoning is a central challenge for large language models (LLMs), requiring not only correct answers but also faithful reasoning processes. Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a promising approach for enhancing such capabilities; however, its ability to foster genuine reasoning remains unclear. We investigate RLVR on two combinatorial problems with fully verifiable solutions: \emph{Activity Scheduling} and the \emph{Longest Increasing Subsequence}, using carefully curated datasets with unique optima. Across multiple reward designs, we find that RLVR improves evaluation metrics but often by reinforcing superficial heuristics rather than acquiring new reasoning strategies. These findings highlight the limits of RLVR generalization, emphasizing the importance of benchmarks that disentangle genuine mathematical reasoning from shortcut exploitation and provide faithful measures of progress. Code available at https://github.com/xashru/rlvr-seq-generalization.

cs.LG↗

arXiv2025

R+R: Revisiting Static Feature-Based Android Malware Detection using Machine Learning

Static feature-based Android malware detection using machine learning (ML) remains critical due to its scalability and efficiency. However, existing approaches often overlook security-critical reproducibility concerns, such as dataset duplication, inadequate hyperparameter tuning, and variance from random initialization. This can significantly compromise the practical effectiveness of these systems. In this paper, we systematically investigate these challenges by proposing a more rigorous methodology for model selection and evaluation. Using two widely used datasets, Drebin and APIGraph, we evaluate six ML models of varying complexity under both offline and continuous active learning settings. Our analysis demonstrates that, contrary to popular belief, well-tuned, simpler models, particularly tree-based methods like XGBoost, consistently outperform more complex neural networks, especially when duplicates are removed. To promote transparency and reproducibility, we open-source our codebase, which is extensible for integrating new models and datasets, facilitating reproducible security research.

cs.CR↗

arXiv2025

Towards Understanding Self-play for LLM Reasoning

Recent advances in large language model (LLM) reasoning, led by reinforcement learning with verifiable rewards (RLVR), have inspired self-play post-training, where models improve by generating and solving their own problems. While self-play has shown strong in-domain and out-of-domain gains, the mechanisms behind these improvements remain poorly understood. In this work, we analyze the training dynamics of self-play through the lens of the Absolute Zero Reasoner, comparing it against RLVR and supervised fine-tuning (SFT). Our study examines parameter update sparsity, entropy dynamics of token distributions, and alternative proposer reward functions. We further connect these dynamics to reasoning performance using pass@k evaluations. Together, our findings clarify how self-play differs from other post-training strategies, highlight its inherent limitations, and point toward future directions for improving LLM math reasoning through self-play.

cs.LG↗

arXiv2025

Adapting Large Language Models to Emerging Cybersecurity using Retrieval Augmented Generation

Security applications are increasingly relying on large language models (LLMs) for cyber threat detection; however, their opaque reasoning often limits trust, particularly in decisions that require domain-specific cybersecurity knowledge. Because security threats evolve rapidly, LLMs must not only recall historical incidents but also adapt to emerging vulnerabilities and attack patterns. Retrieval-Augmented Generation (RAG) has demonstrated effectiveness in general LLM applications, but its potential for cybersecurity remains underexplored. In this work, we introduce a RAG-based framework designed to contextualize cybersecurity data and enhance LLM accuracy in knowledge retention and temporal reasoning. Using external datasets and the Llama-3-8B-Instruct model, we evaluate baseline RAG, an optimized hybrid retrieval approach, and conduct a comparative analysis across multiple performance metrics. Our findings highlight the promise of hybrid retrieval in strengthening the adaptability and reliability of LLMs for cybersecurity tasks.

cs.CR↗

arXiv2025

FACE: Faithful Automatic Concept Extraction

Interpreting deep neural networks through concept-based explanations offers a bridge between low-level features and high-level human-understandable semantics. However, existing automatic concept discovery methods often fail to align these extracted concepts with the model's true decision-making process, thereby compromising explanation faithfulness. In this work, we propose FACE (Faithful Automatic Concept Extraction), a novel framework that augments Non-negative Matrix Factorization (NMF) with a Kullback-Leibler (KL) divergence regularization term to ensure alignment between the model's original and concept-based predictions. Unlike prior methods that operate solely on encoder activations, FACE incorporates classifier supervision during concept learning, enforcing predictive consistency and enabling faithful explanations. We provide theoretical guarantees showing that minimizing the KL divergence bounds the deviation in predictive distributions, thereby promoting faithful local linearity in the learned concept space. Systematic evaluations on ImageNet, COCO, and CelebA datasets demonstrate that FACE outperforms existing methods across faithfulness and sparsity metrics.

cs.CV↗

arXiv2025

Robust ML-based Detection of Conventional, LLM-Generated, and Adversarial Phishing Emails Using Advanced Text Preprocessing

Phishing remains a critical cybersecurity threat, especially with the advent of large language models (LLMs) capable of generating highly convincing malicious content. Unlike earlier phishing attempts which are identifiable by grammatical errors, misspellings, incorrect phrasing, and inconsistent formatting, LLM generated emails are grammatically sound, contextually relevant, and linguistically natural. These advancements make phishing emails increasingly difficult to distinguish from legitimate ones, challenging traditional detection mechanisms. Conventional phishing detection systems often fail when faced with emails crafted by LLMs or manipulated using adversarial perturbation techniques. To address this challenge, we propose a robust phishing email detection system featuring an enhanced text preprocessing pipeline. This pipeline includes spelling correction and word splitting to counteract adversarial modifications and improve detection accuracy. Our approach integrates widely adopted natural language processing (NLP) feature extraction techniques and machine learning algorithms. We evaluate our models on publicly available datasets comprising both phishing and legitimate emails, achieving a detection accuracy of 94.26% and F1-score of 84.39% in model deployment setting. To assess robustness, we further evaluate our models using adversarial phishing samples generated by four attack methods in Python TextAttack framework. Additionally, we evaluate models' performance against phishing emails generated by LLMs including ChatGPT and Llama. Results highlight the resilience of models against evolving AI-powered phishing threats.

cs.CR↗

arXiv2025

Smaller is Better: Enhancing Transparency in Vehicle AI Systems via Pruning

Connected and autonomous vehicles continue to heavily rely on AI systems, where transparency and security are critical for trust and operational safety. Post-hoc explanations provide transparency to these black-box like AI models but the quality and reliability of these explanations is often questioned due to inconsistencies and lack of faithfulness in representing model decisions. This paper systematically examines the impact of three widely used training approaches, namely natural training, adversarial training, and pruning, affect the quality of post-hoc explanations for traffic sign classifiers. Through extensive empirical evaluation, we demonstrate that pruning significantly enhances the comprehensibility and faithfulness of explanations (using saliency maps). Our findings reveal that pruning not only improves model efficiency but also enforces sparsity in learned representation, leading to more interpretable and reliable decisions. Additionally, these insights suggest that pruning is a promising strategy for developing transparent deep learning models, especially in resource-constrained vehicular AI systems.

cs.CV↗

arXiv2025

Do Sparse Subnetworks Exhibit Cognitively Aligned Attention? Effects of Pruning on Saliency Map Fidelity, Sparsity, and Concept Coherence

Prior works have shown that neural networks can be heavily pruned while preserving performance, but the impact of pruning on model interpretability remains unclear. In this work, we investigate how magnitude-based pruning followed by fine-tuning affects both low-level saliency maps and high-level concept representations. Using a ResNet-18 trained on ImageNette, we compare post-hoc explanations from Vanilla Gradients (VG) and Integrated Gradients (IG) across pruning levels, evaluating sparsity and faithfulness. We further apply CRAFT-based concept extraction to track changes in semantic coherence of learned concepts. Our results show that light-to-moderate pruning improves saliency-map focus and faithfulness while retaining distinct, semantically meaningful concepts. In contrast, aggressive pruning merges heterogeneous features, reducing saliency map sparsity and concept coherence despite maintaining accuracy. These findings suggest that while pruning can shape internal representations toward more human-aligned attention patterns, excessive pruning undermines interpretability.

cs.CV↗

arXiv2025

Concept-Based Masking: A Patch-Agnostic Defense Against Adversarial Patch Attacks

Adversarial patch attacks pose a practical threat to deep learning models by forcing targeted misclassifications through localized perturbations, often realized in the physical world. Existing defenses typically assume prior knowledge of patch size or location, limiting their applicability. In this work, we propose a patch-agnostic defense that leverages concept-based explanations to identify and suppress the most influential concept activation vectors, thereby neutralizing patch effects without explicit detection. Evaluated on Imagenette with a ResNet-50, our method achieves higher robust and clean accuracy than the state-of-the-art PatchCleanser, while maintaining strong performance across varying patch sizes and locations. Our results highlight the promise of combining interpretability with robustness and suggest concept-driven defenses as a scalable strategy for securing machine learning models against adversarial patch attacks.

cs.CV↗

arXiv2025

Learning to Look: Cognitive Attention Alignment with Vision-Language Models

Convolutional Neural Networks (CNNs) frequently "cheat" by exploiting superficial correlations, raising concerns about whether they make predictions for the right reasons. Inspired by cognitive science, which highlights the role of attention in robust human perception, recent methods have sought to guide model attention using concept-based supervision and explanation regularization. However, these techniques depend on labor-intensive, expert-provided annotations, limiting their scalability. We propose a scalable framework that leverages vision-language models to automatically generate semantic attention maps using natural language prompts. By introducing an auxiliary loss that aligns CNN attention with these language-guided maps, our approach promotes more reliable and cognitively plausible decision-making without manual annotation. Experiments on challenging datasets, ColoredMNIST and DecoyMNIST, show that our method achieves state-of-the-art performance on ColorMNIST and remains competitive with annotation-heavy baselines on DecoyMNIST, demonstrating improved generalization, reduced shortcut reliance, and model attention that better reflects human intuition.

cs.CV↗

arXiv2025

ADAPT: A Pseudo-labeling Approach to Combat Concept Drift in Malware Detection

Machine learning models are commonly used for malware classification; however, they suffer from performance degradation over time due to concept drift. Adapting these models to changing data distributions requires frequent updates, which rely on costly ground truth annotations. While active learning can reduce the annotation burden, leveraging unlabeled data through semi-supervised learning remains a relatively underexplored approach in the context of malware detection. In this research, we introduce \texttt{ADAPT}, a novel pseudo-labeling semi-supervised algorithm for addressing concept drift. Our model-agnostic method can be applied to various machine learning models, including neural networks and tree-based algorithms. We conduct extensive experiments on five diverse malware detection datasets spanning Android, Windows, and PDF domains. The results demonstrate that our method consistently outperforms baseline models and competitive benchmarks. This work paves the way for more effective adaptation of machine learning models to concept drift in malware detection.

cs.LG↗

arXiv2025

Too Much to Trust? Measuring the Security and Cognitive Impacts of Explainability in AI-Driven SOCs

Explainable AI (XAI) holds significant promise for enhancing the transparency and trustworthiness of AI-driven threat detection in Security Operations Centers (SOCs). However, identifying the appropriate level and format of explanation, particularly in environments that demand rapid decision-making under high-stakes conditions, remains a complex and underexplored challenge. To address this gap, we conducted a three-month mixed-methods study combining an online survey (N1=248) with in-depth interviews (N2=24) to examine (1) how SOC analysts conceptualize AI-generated explanations and (2) which types of explanations are perceived as actionable and trustworthy across different analyst roles. Our findings reveal that participants were consistently willing to accept XAI outputs, even in cases of lower predictive accuracy, when explanations were perceived as relevant and evidence-backed. Analysts repeatedly emphasized the importance of understanding the rationale behind AI decisions, expressing a strong preference for contextual depth over a mere presentation of outcomes on dashboards. Building on these insights, this study re-evaluates current explanation methods within security contexts and demonstrates that role-aware, context-rich XAI designs aligned with SOC workflows can substantially improve practical utility. Such tailored explainability enhances analyst comprehension, increases triage efficiency, and supports more confident responses to evolving threats.

cs.CR↗