Publications

1. NeurIPS 2024

   Spotlight

   CTIBench: A Benchmark for Evaluating LLMs in Cyber Threat Intelligence

   Md Tanvirul Alam, Dipkamal Bhusal, Le Nguyen, and Nidhi Rastogi

   In Advances in Neural Information Processing Systems, 2024

   Datasets and Benchmarks Track

   Abstract arXiv Code

   CTIBench introduces a task-focused benchmark for cyber threat intelligence that probes factual recall, applied reasoning, and robustness, showing that even strong language models remain unreliable on practical CTI workflows.

2. CVPR 2026

   Findings

   SPHINX: A Synthetic Environment for Visual Perception and Reasoning

   Md Tanvirul Alam, Saksham Aggarwal, Justin Yang Chae, and Nidhi Rastogi

   In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2026

   Abstract arXiv Code

   SPHINX builds a synthetic visual reasoning environment with procedurally generated puzzles and verifiable answers, enabling precise evaluation of multimodal reasoning and showing that RLVR can improve large vision-language models on these tasks.

3. arXiv 2026

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

   Md Tanvirul Alam, Aritran Piplai, Ionut Cardei, Nidhi Rastogi, and Peter J. Worth Jr

   arXiv preprint, 2026

   Abstract arXiv

   Minerva studies RL with verifiable rewards for cyber threat intelligence tasks, using structured CTI outputs and deterministic verifiers to improve model accuracy and robustness beyond supervised fine-tuning.

4. RAID 2023

   Looking Beyond IoCs: Automatically Extracting Attack Patterns from External CTI

   Md Tanvirul Alam, Dipkamal Bhusal, Youngja Park, and Nidhi Rastogi

   In International Symposium on Research in Attacks, Intrusions and Defenses, 2023

   Abstract arXiv

   Looking Beyond IoCs introduces LADDER, a framework for extracting text-based attack patterns from CTI reports and mapping them to MITRE ATT&CK so analysts can reason about evolving threats beyond brittle indicators.

2025

1. WAITI 2025

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

   Md Tanvirul Alam, Dipkamal Bhusal, Salman Ahmad, Nidhi Rastogi, and Peter Worth

   WAITI Workshop, 2025

   Abstract arXiv Code

   AthenaBench extends CTIBench with a stronger data pipeline, refined metrics, and a new risk-mitigation task, revealing that current frontier models still struggle on reasoning-heavy CTI problems.

2. MATH-AI 2025

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

   Md Tanvirul Alam and Nidhi Rastogi

   MATH-AI Workshop, 2025

   Abstract arXiv Code

   This paper tests RLVR on combinatorial reasoning problems with fully verifiable solutions and finds that better scores often come from exploiting shortcuts rather than learning genuinely transferable reasoning strategies.

3. ACSAC 2025

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

   Md Tanvirul Alam, Dipkamal Bhusal, and Nidhi Rastogi

   Proceedings of the Annual Computer Security Applications Conference, 2025

   Abstract arXiv

   R+R revisits static feature-based Android malware detection with modern experimental controls, showing where traditional machine learning pipelines still work well and where they fail under realistic distribution shifts.

4. RAID 2025

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

   Md Tanvirul Alam, Aritran Piplai, and Nidhi Rastogi

   International Symposium on Research in Attacks, Intrusions and Defenses, 2025

   Abstract arXiv

   ADAPT tackles concept drift in malware detection by using pseudo-labeling to refresh models on shifting data distributions, improving robustness when labeled updates are scarce.

5. MATH-AI 2025

   Towards Understanding Self-play for LLM Reasoning

   Justin Yang Chae, Md Tanvirul Alam, and Nidhi Rastogi

   MATH-AI Workshop, 2025

   Abstract arXiv

   This work studies how self-play improves language-model reasoning by analyzing the problems generated during training and the mechanisms that help reasoning ability emerge without external supervision.

2024

1. ACSAC 2024

   SECURE: Benchmarking Large Language Models for Cybersecurity

   Dipkamal Bhusal, Md Tanvirul Alam, Le Nguyen, Ashim Mahara, Zachary Lightcap, Rodney Frazier, Romy Fieblinger, Grace Long Torales, Benjamin A. Blakely, and Nidhi Rastogi

   Proceedings of the Annual Computer Security Applications Conference, 2024

   Abstract arXiv

   SECURE introduces a broad cybersecurity benchmark spanning knowledge extraction, understanding, and reasoning tasks, and shows that current large language models remain inconsistent on applied cyber problems.

2. EuroS&PW 2024

   Actionable Cyber Threat Intelligence using Knowledge Graphs and Large Language Models

   Romy Fieblinger, Md Tanvirul Alam, and Nidhi Rastogi

   In IEEE European Symposium on Security and Privacy Workshops, 2024

   Abstract arXiv

   This paper combines open-source language models with knowledge-graph construction to extract actionable cyber threat intelligence from unstructured reports and study how well structured CTI can be built at scale.

3. EuroS&P 2024

   PASA: Attack Agnostic Unsupervised Adversarial Detection using Prediction and Attribution Sensitivity Analysis

   Dipkamal Bhusal, Md Tanvirul Alam, Monish Kumar Manikya Veerabhadran, Michael Clifford, Sara Rampazzi, and Nidhi Rastogi

   IEEE European Symposium on Security and Privacy, 2024

   Abstract arXiv

   PASA detects adversarial examples without attack-specific tuning by measuring how unstable both predictions and feature attributions become under noise, using thresholds learned only from benign data.

2022

1. arXiv 2022

   CyNER: A Python Library for Cybersecurity Named Entity Recognition

   Md Tanvirul Alam, Dipkamal Bhusal, Youngja Park, and Nidhi Rastogi

   arXiv preprint, 2022

   Abstract arXiv Code

   CyNER packages transformer models, heuristic extractors, and generic NER components into an open-source library for extracting cybersecurity entities and indicators of compromise from heterogeneous threat reports.

2020

1. arXiv 2020

   Bangla Text Classification using Transformers

   Tanvirul Alam, Akib Khan, and Firoj Alam

   arXiv preprint, 2020

   Abstract arXiv Code

   This paper fine-tunes multilingual transformer models for Bangla sentiment analysis, emotion detection, news categorization, and authorship attribution, reporting state-of-the-art gains across six benchmark datasets.

2. W-NUT 2020

   Punctuation Restoration using Transformer Models for High-and Low-Resource Languages

   Tanvirul Alam, Akib Khan, and Firoj Alam

   In Proceedings of the 6th Workshop on Noisy User-generated Text, 2020

   Abstract PDF Code

   This paper studies punctuation restoration for both high-resource and low-resource settings, explores transformer-based models for ASR post-processing, and introduces an augmentation strategy that improves robustness while establishing a public baseline for Bangla.

3. SPECOM 2020

   Lightweight CNN for Robust Voice Activity Detection

   Tanvirul Alam and Akib Khan

   In International Conference on Speech and Computer, 2020

   Abstract Code

   Voice activity detection is a critical preprocessing step for speech systems. This work introduces a lightweight CNN architecture and improves noisy-condition robustness with strong regularization and knowledge distillation, yielding large relative EER gains over a parameter-matched DNN baseline.