Improving LLM Agents with Reinforcement Learning on Cryptographic CTF Challenges

• URL: http://arxiv.org/abs/2506.02048v2
• Date: Sun, 17 Aug 2025 22:28:50 GMT
• Title: Improving LLM Agents with Reinforcement Learning on Cryptographic CTF Challenges
• Authors: Lajos Muzsai, David Imolai, András Lukács,
• Abstract summary: 'Random-Crypto' is a procedurally generated cryptographic dataset designed to unlock the potential of Reinforcement Learning.<n>We fine-tune a Python tool-augmented Llama-3.1-8B via Group Relative Policy Optimization.<n>The resulting agent achieves a significant improvement in Pass@8 on previously unseen challenges.
• Score: 0.0
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: We present 'Random-Crypto', a procedurally generated cryptographic Capture The Flag (CTF) dataset designed to unlock the potential of Reinforcement Learning (RL) for LLM-based agents in security-sensitive domains. Cryptographic reasoning offers an ideal RL testbed: it combines precise validation, structured multi-step inference, and reliance on reliable computational tool use. Leveraging these properties, we fine-tune a Python tool-augmented Llama-3.1-8B via Group Relative Policy Optimization (GRPO) in a secure execution environment. The resulting agent achieves a significant improvement in Pass@8 on previously unseen challenges. Moreover, the improvements generalize to two external benchmarks: 'picoCTF', spanning both crypto and non-crypto tasks, and 'AICrypto MCQ', a multiple-choice benchmark of 135 cryptography questions. Ablation studies attribute the gains to enhanced tool usage and procedural reasoning. These findings position 'Random-Crypto' as a rich training ground for building intelligent, adaptable LLM agents capable of handling complex cybersecurity tasks.

Related papers

• CIPHER: Cryptographic Insecurity Profiling via Hybrid Evaluation of Responses [0.0]
  We introduce CIPHER, a benchmark for measuring cryptographic vulnerability incidence in Python code.<n> CIPHER uses insecure/neutral/secure prompt variants per task, a cryptography-specific vulnerability taxonomy, and line-level attribution.<n>We find that explicit secure prompting reduces some targeted issues but does not reliably eliminate cryptographic vulnerabilities overall.
  arXiv  Detail & Related papers  (2026-02-01T21:06:54Z)
• CVeDRL: An Efficient Code Verifier via Difficulty-aware Reinforcement Learning [57.24524263804788]
  Code verifiers play a critical role in post-verification for LLM-based code generation.<n>Existing supervised fine-tuning methods suffer from data scarcity, high failure rates, and poor inference efficiency.<n>We show that naive RL with only functionality rewards fails to generate effective unit tests for difficult branches and samples.
  arXiv  Detail & Related papers  (2026-01-30T10:33:29Z)
• CryptoQA: A Large-scale Question-answering Dataset for AI-assisted Cryptography [13.643089244089873]
  We present CryptoQA, the first large-scale question-answering dataset specifically designed for cryptography.<n>We benchmark 15 state-of-the-art LLMs on CryptoQA, evaluating their factual accuracy, mathematical reasoning, consistency, referencing, and robustness to adversarial samples.<n>Our results reveal significant performance deficits of LLMs, particularly on tasks that require formal reasoning and precise mathematical knowledge.
  arXiv  Detail & Related papers  (2025-12-02T10:35:36Z)
• Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training [121.5858973157225]
  We investigate the effects of prolonged reinforcement learning on a small language model across a diverse set of reasoning domains.<n>We introduce controlled KL regularization, clipping ratio, and periodic reference policy resets as critical components for unlocking long-term performance gains.<n>Our model achieves significant improvements over strong baselines, including +14.7% on math, +13.9% on coding, and +54.8% on logic puzzle tasks.
  arXiv  Detail & Related papers  (2025-07-16T17:59:24Z)
• AICrypto: A Comprehensive Benchmark For Evaluating Cryptography Capabilities of Large Language Models [18.222665072623755]
  Large language models (LLMs) have demonstrated remarkable capabilities across a variety of domains.<n>Our benchmark comprises 135 multiple-choice questions, 150 capture-the-flag (CTF) challenges, and 18 proof problems.<n>Our evaluation of 17 leading LLMs reveals that state-of-the-art models match or even surpass human experts in memorizing cryptographic concepts.
  arXiv  Detail & Related papers  (2025-07-13T11:11:01Z)
• Simplicity by Obfuscation: Evaluating LLM-Driven Code Transformation with Semantic Elasticity [4.458584890504334]
  Code obfuscation aims to prevent reverse engineering and intellectual property theft.<n>The recent development of large language models paves the way for practical applications in different domains.<n>This work performs an empirical study on the ability of LLMs to obfuscate Python source code.
  arXiv  Detail & Related papers  (2025-04-18T18:29:23Z)
• TFHE-Coder: Evaluating LLM-agentic Fully Homomorphic Encryption Code Generation [10.597643264309415]
  Homomorphic Encryption over the torus (TFHE) enables encrypted computation on data without decryption.<n>Despite its potential in privacy preserving machine learning, secure multi party computation, private blockchain transactions, and secure medical diagnostics, its adoption remains limited due to cryptographic complexity and usability challenges.<n>This work establishes the first benchmark for TFHE code generation, demonstrating how LLMs, when augmented with domain-specific feedback, can bridge the expertise gap in FHE code generation.
  arXiv  Detail & Related papers  (2025-03-15T17:57:44Z)
• Cryptanalysis via Machine Learning Based Information Theoretic Metrics [58.96805474751668]
  We propose two novel applications of machine learning (ML) algorithms to perform cryptanalysis on any cryptosystem.<n>These algorithms can be readily applied in an audit setting to evaluate the robustness of a cryptosystem.<n>We show that our classification model correctly identifies the encryption schemes that are not IND-CPA secure, such as DES, RSA, and AES ECB, with high accuracy.
  arXiv  Detail & Related papers  (2025-01-25T04:53:36Z)
• QuanCrypt-FL: Quantized Homomorphic Encryption with Pruning for Secure Federated Learning [0.48342038441006796]
  We propose QuanCrypt-FL, a novel algorithm that combines low-bit quantization and pruning techniques to enhance protection against attacks. We validate our approach on MNIST, CIFAR-10, and CIFAR-100 datasets, demonstrating superior performance compared to state-of-the-art methods. QuanCrypt-FL achieves up to 9x faster encryption, 16x faster decryption, and 1.5x faster inference compared to BatchCrypt, with training time reduced by up to 3x.
  arXiv  Detail & Related papers  (2024-11-08T01:46:00Z)
• Language Models are Hidden Reasoners: Unlocking Latent Reasoning Capabilities via Self-Rewarding [74.31981011985681]
  Large language models (LLMs) have shown impressive capabilities, but still struggle with complex reasoning tasks requiring multiple steps. We introduce LaTent Reasoning Optimization (LaTRO), a principled framework that formulates reasoning as sampling from a latent distribution. We validate LaTRO through experiments on GSM8K and ARC-Challenge datasets using multiple model architectures.
  arXiv  Detail & Related papers  (2024-11-06T22:02:30Z)
• From Solitary Directives to Interactive Encouragement! LLM Secure Code Generation by Natural Language Prompting [24.27542373791212]
  This work introduces SecCode, a framework that leverages an innovative interactive encouragement prompting (EP) technique for secure code generation with textitonly NL prompts. SecCode functions through three stages: 1) Code Generation using NL Prompts; 2) Code Vulnerability Detection and Fixing, utilising our proposed encouragement prompting; 3) Vulnerability Cross-Checking and Code Security Refinement.
  arXiv  Detail & Related papers  (2024-10-18T09:32:08Z)
• HexaCoder: Secure Code Generation via Oracle-Guided Synthetic Training Data [60.75578581719921]
  Large language models (LLMs) have shown great potential for automatic code generation. Recent studies highlight that many LLM-generated code contains serious security vulnerabilities. We introduce HexaCoder, a novel approach to enhance the ability of LLMs to generate secure codes.
  arXiv  Detail & Related papers  (2024-09-10T12:01:43Z)
• ShadowCode: Towards (Automatic) External Prompt Injection Attack against Code LLMs [56.46702494338318]
  This paper introduces a new attack paradigm: (automatic) external prompt injection against code-oriented large language models.<n>We propose ShadowCode, a simple yet effective method that automatically generates induced perturbations based on code simulation.<n>We evaluate our method across 13 distinct malicious objectives, generating 31 threat cases spanning three popular programming languages.
  arXiv  Detail & Related papers  (2024-07-12T10:59:32Z)
• FoC: Figure out the Cryptographic Functions in Stripped Binaries with LLMs [51.898805184427545]
  We propose a novel framework called FoC to Figure out the Cryptographic functions in stripped binaries.<n>We first build a binary large language model (FoC-BinLLM) to summarize the semantics of cryptographic functions in natural language.<n>We then build a binary code similarity model (FoC-Sim) upon the FoC-BinLLM to create change-sensitive representations and use it to retrieve similar implementations of unknown cryptographic functions in a database.
  arXiv  Detail & Related papers  (2024-03-27T09:45:33Z)
• CodeChameleon: Personalized Encryption Framework for Jailbreaking Large Language Models [49.60006012946767]
  We propose CodeChameleon, a novel jailbreak framework based on personalized encryption tactics. We conduct extensive experiments on 7 Large Language Models, achieving state-of-the-art average Attack Success Rate (ASR) Remarkably, our method achieves an 86.6% ASR on GPT-4-1106.
  arXiv  Detail & Related papers  (2024-02-26T16:35:59Z)
• Online Safety Property Collection and Refinement for Safe Deep Reinforcement Learning in Mapless Navigation [79.89605349842569]
  We introduce the Collection and Refinement of Online Properties (CROP) framework to design properties at training time. CROP employs a cost signal to identify unsafe interactions and use them to shape safety properties. We evaluate our approach in several robotic mapless navigation tasks and demonstrate that the violation metric computed with CROP allows higher returns and lower violations over previous Safe DRL approaches.
  arXiv  Detail & Related papers  (2023-02-13T21:19:36Z)
• THE-X: Privacy-Preserving Transformer Inference with Homomorphic Encryption [112.02441503951297]
  Privacy-preserving inference of transformer models is on the demand of cloud service users. We introduce $textitTHE-X$, an approximation approach for transformers, which enables privacy-preserving inference of pre-trained models.
  arXiv  Detail & Related papers  (2022-06-01T03:49:18Z)
• Robust Deep Reinforcement Learning through Adversarial Loss [74.20501663956604]
  Recent studies have shown that deep reinforcement learning agents are vulnerable to small adversarial perturbations on the agent's inputs. We propose RADIAL-RL, a principled framework to train reinforcement learning agents with improved robustness against adversarial attacks.
  arXiv  Detail & Related papers  (2020-08-05T07:49:42Z)
• Faster Secure Data Mining via Distributed Homomorphic Encryption [108.77460689459247]
  Homomorphic Encryption (HE) is receiving more and more attention recently for its capability to do computations over the encrypted field. We propose a novel general distributed HE-based data mining framework towards one step of solving the scaling problem. We verify the efficiency and effectiveness of our new framework by testing over various data mining algorithms and benchmark data-sets.
  arXiv  Detail & Related papers  (2020-06-17T18:14:30Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.