pith. machine review for the scientific record. sign in

arxiv: 2604.05440 · v1 · submitted 2026-04-07 · 💻 cs.CR · cs.AI

Recognition: 2 theorem links

· Lean Theorem

LanG -- A Governance-Aware Agentic AI Platform for Unified Security Operations

Anes Abdennebi , Nadjia Kara , Laaziz Lahlou , Hakima Ould-Slimane
Authors on Pith no claims yet

Pith reviewed 2026-05-10 19:47 UTC · model grok-4.3

classification 💻 cs.CR cs.AI
keywords agentic AIsecurity operationsLLM rule generationgovernance guardrailsintrusion detectionSOC platformopen-source securityattack reconstruction
0
0 comments X

The pith

LanG combines multiple AI security components into one open-source platform that enforces governance policies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces LanG to address alert fatigue and fragmented tools in security operations by integrating a correlation engine for unified incident records, an agentic orchestrator with human checkpoints, an LLM-based rule generator for IDS signatures, a three-phase attack reconstructor, and a governance layer. Benchmarks show 87% F1 on event correlation, 96.2% rule acceptance, 87.5% kill-chain accuracy, and 98.1% F1 on guardrails with reported zero false positives, plus strong results on anomaly detection at low latency. A comparison against eight other SOC platforms claims LanG is the only open-source option that satisfies multiple industrial capabilities while supporting local multi-tenant deployment. Sympathetic readers would see value in a single tool that adds AI assistance without losing policy control.

Core claim

LanG is an open-source governance-aware agentic AI platform for unified security operations that delivers a Unified Incident Context Record with correlation engine, a LangGraph-based orchestrator with human-in-the-loop checkpoints, a fine-tuned LLM rule generator for Snort/Suricata/YARA rules, a three-phase attack reconstructor using Louvain detection and Bayesian scoring, and a layered architecture exposing tools via Model Context Protocol under an AI Governance Policy Engine with a two-layer guardrail pipeline, achieving the listed performance metrics and uniquely meeting multiple SOC requirements in one deployable system.

What carries the argument

The layered Governance-MCP-Agentic AI-Security architecture that governs all components through an AI Governance Policy Engine and a two-layer guardrail pipeline of regex plus semantic classifier.

If this is right

  • Generates deployable security rules with 96.2% average acceptance rate across four base models.
  • Correlates cross-source events into unified records at 87% F1 score.
  • Reconstructs attack kill-chains at 87.5% accuracy through community detection and Bayesian hypothesis scoring.
  • Detects anomalies at 99% weighted F1 and threats at 91% weighted F1 with inference around 21 ms.
  • Supports fully local, multi-tenant, role-based deployment for managed security service providers.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The human-in-the-loop design could lower barriers to regulated adoption by keeping oversight on AI actions.
  • If the guardrails prove robust beyond the tested conditions, the platform could serve as a template for governed AI in other operational domains.
  • Open-source availability might enable community testing on diverse network environments to strengthen generalization claims.
  • Fast local inference times open the possibility of embedding the system directly on edge security devices.

Load-bearing premise

The reported accuracy, acceptance, and zero false-positive figures will continue to hold when the system encounters real-world adversarial inputs or previously unseen attack patterns.

What would settle it

Deploy the guardrail pipeline and rule generator against a curated set of adversarial prompts or novel attack traces and measure whether false positives appear or rule acceptance falls substantially below 96%.

Figures

Figures reproduced from arXiv: 2604.05440 by Anes Abdennebi, Hakima Ould-Slimane, Laaziz Lahlou, Nadjia Kara.

Figure 1
Figure 1. Figure 1: LanG platform layered architecture. The Security layer provides guardrails that protect the Agentic AI agents, their capabilities are exposed via MCP tools and resources, and all MCP access is governed by the top-level governance policy. Arrows indicate the dependency direction (bottom-up). III. SYSTEM ARCHITECTURE [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: SOC Agentic Pipeline: five agent nodes with two [PITH_FULL_IMAGE:figures/full_fig_p011_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Governance enforcement architecture of the LanG platform. [PITH_FULL_IMAGE:figures/full_fig_p018_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: MSSP deployment topology. The central LanG instance [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Syntactic validity rate (%) by model and rule format on [PITH_FULL_IMAGE:figures/full_fig_p019_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Ablation: effect of the RulePostProcessor on syntactic validity rate. Post-processing improves validity by 3.2–5.1 percentage points, with the largest gain on Phi-3-mini. TABLE XIII: Correlation Engine Grouping and Attack Recon￾structor Performance Correlation Reconstructor Scenario Prec. Rec. F1 Chain Bayes. (%) (%) (%) Acc. (%) Score BruteForce 91.0 88.0 89.0 91 0.82 Exfiltration 87.0 84.0 86.0 86 0.78 W… view at source ↗
Figure 7
Figure 7. Figure 7: Guardrail recall per adversarial category: Layer 1 (regex [PITH_FULL_IMAGE:figures/full_fig_p025_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Reconstruction time vs. number of input events. Phase 2 [PITH_FULL_IMAGE:figures/full_fig_p026_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Simplified attack graph for the brute-force-to [PITH_FULL_IMAGE:figures/full_fig_p034_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Example attack scenario visualization generated by Phase 3. Nodes are labelled by kill-chain phase (e.g., reconnaissance, [PITH_FULL_IMAGE:figures/full_fig_p035_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Example of a reconstructed cyberattack scenario after being flagged then acted against (through a newly generated [PITH_FULL_IMAGE:figures/full_fig_p035_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Example of agentic AI flow for anomaly & threat detection, log analysis, and new detection rule generation of a [PITH_FULL_IMAGE:figures/full_fig_p036_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Example of a threat path where the threat detection agent communicates with the rule generation agent, then awaiting [PITH_FULL_IMAGE:figures/full_fig_p036_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Example of a the generated rules using one of the finetuned models (Qwen3-1.7-base). [PITH_FULL_IMAGE:figures/full_fig_p037_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Example of a complete agentic pipeline flow of a detected threat and generated then deployed its respective detection [PITH_FULL_IMAGE:figures/full_fig_p037_15.png] view at source ↗
read the original abstract

Modern Security Operations Centers struggle with alert fatigue, fragmented tooling, and limited cross-source event correlation. Challenges that current Security Information Event Management and Extended Detection and Response systems only partially address through fragmented tools. This paper presents the LLM-assisted network Governance (LanG), an open-source, governance-aware agentic AI platform for unified security operations contributing: (i) a Unified Incident Context Record with a correlation engine (F1 = 87%), (ii) an Agentic AI Orchestrator on LangGraph with human-in-the-loop checkpoints, (iii) an LLM-based Rule Generator finetuned on four base models producing deployable Snort 2/3, Suricata, and YARA rules (average acceptance rate 96.2%), (iv) a Three-Phase Attack Reconstructor combining Louvain community detection, LLM-driven hypothesis generation, and Bayesian scoring (87.5% kill-chain accuracy), and (v) a layered Governance-MCP-Agentic AI-Security architecture where all tools are exposed via the Model Context Protocol, governed by an AI Governance Policy Engine with a two-layer guardrail pipeline (regex + Llama Prompt Guard 2 semantic classifier, achieving 98.1% F1 score with experimental zero false positives). Designed for Managed Security Service Providers, the platform supports multi-tenant isolation, role-based access, and fully local deployment. Finetuned anomaly and threat detectors achieve weighted F1 scores of 99.0% and 91.0%, respectively, in intrusion-detection benchmarks, running inferences in $\approx$21 ms with a machine-side mean time to detect of 1.58 s, and the rule generator exceeds 91% deployability on live IDS engines. A systematic comparison against eight SOC platforms confirms that LanG uniquely satisfies multiple industrial capabilities all in one open-source tool, while enforcing selected AI governance policies.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript presents LanG, an open-source governance-aware agentic AI platform for unified security operations addressing SOC challenges like alert fatigue and fragmented tooling. It contributes (i) a Unified Incident Context Record with correlation engine (F1=87%), (ii) an Agentic AI Orchestrator on LangGraph with human-in-the-loop checkpoints, (iii) an LLM-based Rule Generator fine-tuned on four base models yielding 96.2% average acceptance rate for Snort/Suricata/YARA rules, (iv) a Three-Phase Attack Reconstructor using Louvain detection, LLM hypothesis generation, and Bayesian scoring (87.5% kill-chain accuracy), and (v) a layered Governance-MCP-Agentic AI-Security architecture with an AI Governance Policy Engine featuring a two-layer guardrail (regex + Llama Prompt Guard 2) achieving 98.1% F1 with experimental zero false positives. Additional elements include fine-tuned anomaly (99.0% F1) and threat (91.0% F1) detectors with ~21 ms inference and 1.58 s mean time to detect, plus a systematic comparison against eight SOC platforms asserting that LanG uniquely combines multiple industrial capabilities in one open-source tool while enforcing selected AI governance policies. The platform supports multi-tenant isolation, role-based access, and fully local deployment.

Significance. If the reported empirical results are validated with full experimental details, the work would offer a meaningful contribution as an integrated open-source platform that unifies correlation, orchestration, rule generation, attack reconstruction, and governance enforcement for SOC/MSSP use. The open-source availability, emphasis on local deployment, human-in-the-loop design, and explicit comparison to existing platforms are positive aspects that could aid adoption and reproducibility. The governance layer via the two-layer guardrail and Model Context Protocol exposure represents a timely focus on AI safety in security tooling. However, the significance depends heavily on whether the performance claims generalize beyond the (unspecified) test conditions.

major comments (3)
  1. Abstract: The manuscript reports multiple quantitative results (correlation F1=87%, rule acceptance 96.2%, attack reconstruction 87.5%, guardrail F1=98.1% with zero false positives, anomaly/threat detector F1 scores of 99.0%/91.0%) but supplies no information on datasets, baselines, evaluation protocols, statistical tests, or ablation studies. This omission is load-bearing for the central claims of effectiveness and uniqueness, as it is impossible to determine whether the numbers support the stated conclusions or generalize to real SOC conditions.
  2. Abstract (governance component): The two-layer guardrail (regex + Llama Prompt Guard 2) is presented as achieving 98.1% F1 'with experimental zero false positives' and is central to the claim that LanG 'uniquely' enforces AI governance policies relative to the eight compared platforms. No details are given on the test inputs (e.g., whether they include adversarial or novel prompts relevant to rule generation and agent actions), so the zero-FP result does not demonstrably transfer to the conditions that matter for the uniqueness assertion.
  3. Abstract (comparison): The assertion that 'a systematic comparison against eight SOC platforms confirms that LanG uniquely satisfies multiple industrial capabilities all in one open-source tool' is load-bearing for the paper's primary contribution claim, yet no information is provided on the identity of the eight platforms, the precise capabilities evaluated, or the evaluation methodology. Without this, the uniqueness conclusion cannot be assessed.
minor comments (2)
  1. Abstract: The phrase 'finetuned on four base models' is used without naming the models or describing the fine-tuning data, hyperparameters, or evaluation split, which reduces clarity even though it is not central to the main claims.
  2. Abstract: The inference time ('≈21 ms') and mean time to detect (1.58 s) are reported without specifying hardware, batch size, or test conditions, which is a minor presentation issue for reproducibility.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed and constructive review. The comments correctly identify areas where the abstract could better support the quantitative claims and contribution statements. We have revised the manuscript to address these points by expanding the abstract and adding supporting details in the main text.

read point-by-point responses

  1. Referee: Abstract: The manuscript reports multiple quantitative results (correlation F1=87%, rule acceptance 96.2%, attack reconstruction 87.5%, guardrail F1=98.1% with zero false positives, anomaly/threat detector F1 scores of 99.0%/91.0%) but supplies no information on datasets, baselines, evaluation protocols, statistical tests, or ablation studies. This omission is load-bearing for the central claims of effectiveness and uniqueness, as it is impossible to determine whether the numbers support the stated conclusions or generalize to real SOC conditions.

    Authors: We agree that the abstract would be strengthened by including high-level experimental context. In the revised manuscript we have expanded the abstract to reference the primary datasets and benchmarks used for each component, the evaluation protocols (including hold-out testing and cross-validation), and the baselines. We have also added a concise summary of ablation studies and statistical significance testing in the experimental sections to better substantiate the reported metrics. revision: yes

  2. Referee: Abstract (governance component): The two-layer guardrail (regex + Llama Prompt Guard 2) is presented as achieving 98.1% F1 'with experimental zero false positives' and is central to the claim that LanG 'uniquely' enforces AI governance policies relative to the eight compared platforms. No details are given on the test inputs (e.g., whether they include adversarial or novel prompts relevant to rule generation and agent actions), so the zero-FP result does not demonstrably transfer to the conditions that matter for the uniqueness assertion.

    Authors: We accept that additional information on the guardrail test inputs is required to support the zero false-positive claim under relevant conditions. The revised abstract now notes that the evaluation included adversarial and domain-specific prompts. We have also added a paragraph in the governance section describing the test set composition, including the proportion of adversarial and novel prompts related to rule generation and agent actions, along with the exact evaluation protocol. revision: yes

  3. Referee: Abstract (comparison): The assertion that 'a systematic comparison against eight SOC platforms confirms that LanG uniquely satisfies multiple industrial capabilities all in one open-source tool' is load-bearing for the paper's primary contribution claim, yet no information is provided on the identity of the eight platforms, the precise capabilities evaluated, or the evaluation methodology. Without this, the uniqueness conclusion cannot be assessed.

    Authors: We acknowledge that the abstract's uniqueness claim would be clearer with a brief indication of the comparison scope. The detailed comparison, including platform identities, evaluated capabilities, and methodology, appears in Section 6. In the revision we have updated the abstract to name the platforms and summarize the capability categories and evaluation approach, while retaining the full analysis in the main text. revision: partial

Circularity Check

0 steps flagged

No circularity: all reported results are direct empirical measurements with no derivations or self-referential definitions.

full rationale

The paper presents an agentic AI platform and reports performance via benchmark F1 scores, acceptance rates, accuracy figures, and a platform comparison. These are framed as direct experimental outcomes from specific test conditions rather than quantities derived from other fitted parameters or prior self-citations. No equations, ansatzes, uniqueness theorems, or load-bearing self-citations appear in the provided text. The central claims rest on empirical validation and open-source implementation details, remaining independent of any circular reduction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The abstract contains no mathematical derivations, free parameters, or postulated entities. All contributions are engineering components and reported performance figures.

pith-pipeline@v0.9.0 · 5657 in / 1259 out tokens · 84361 ms · 2026-05-10T19:47:23.534849+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Decision Evidence Maturity Model for Agentic AI: A Property-Level Method Specification

    cs.CY 2026-04 unverdicted novelty 4.0

    DEMM defines four executable evidence-sufficiency categories plus a conflicting category for agentic AI decisions and rolls per-property verdicts into a five-level maturity rubric.

Reference graph

Works this paper leans on

119 extracted references · 20 canonical work pages · cited by 1 Pith paper · 5 internal anchors

  1. [1]

    A survey on iot security: Application areas, security threats, and solution architectures,

    V . Hassija, V . Chamola, V . Saxena, D. Jain, P. Goyal, and B. Sikdar, “A survey on iot security: Application areas, security threats, and solution architectures,”IEEE Access, vol. 7, pp. 82 721–82 743, 2019

    2019

  2. [2]

    Demystifying iot security: An exhaustive survey on iot vulnerabilities and a first empirical look on internet-scale iot exploitations,

    N. Neshenko, E. Bou-Harb, J. Crichigno, G. Kaddoum, and N. Ghani, “Demystifying iot security: An exhaustive survey on iot vulnerabilities and a first empirical look on internet-scale iot exploitations,”IEEE Communications Surveys & Tutorials, vol. 21, no. 3, pp. 2702–2733, 2019

    2019

  3. [3]

    Security for 5g and beyond,

    I. Ahmad, S. Shahabuddin, T. Kumar, J. Okwuibe, A. Gurtov, and M. Ylianttila, “Security for 5g and beyond,”IEEE Communications Surveys & Tutorials, vol. 21, no. 4, pp. 3682–3722, 2019

    2019

  4. [4]

    Market guide for extended detection and response,

    Gartner, “Market guide for extended detection and response,” Gartner Research Report, 2022, accessed: 2026-03-01

    2022

  5. [5]

    A tale of three security operation centers,

    S. C. Sundaramurthy, J. Case, T. Truong, L. Zomlot, and M. Hoffmann, “A tale of three security operation centers,” inProceedings of the 2014 ACM Workshop on Security Information Workers, ser. SIW ’14. New York, NY , USA: Association for Computing Machinery, 2014, p. 43–50. [Online]. Available: https://doi.org/10.1145/2663887.2663904

    work page doi:10.1145/2663887.2663904 2014

  6. [6]

    Security information and event management (siem): Analysis, trends, and usage in critical infrastructures,

    G. González-Granadillo, S. González-Zarzosa, and R. Diaz, “Security information and event management (siem): Analysis, trends, and usage in critical infrastructures,”Sensors, vol. 21, no. 14, 2021. [Online]. Available: https://www.mdpi.com/1424-8220/21/14/4759

    2021

  7. [8]

    The voice of the SOC ana- lyst report,

    Trellix (formerly FireEye/Mandiant), “The voice of the SOC ana- lyst report,” https://www.trellix.com/about/newsroom/stories/research/ xdr-solving-the-soc-analyst-alert-fatigue-problem/, 2023, accessed: 2026-03-01

    2023

  8. [9]

    Matched and mismatched SOCs: A qualitative study on security operations center issues,

    F. B. Kokulu, A. Soneji, T. Bao, Y . Shoshitaishvili, Z. Zhao, A. Doupé, and G.-J. Ahn, “Matched and mismatched socs: A qualitative study on security operations center issues,” ser. CCS ’19. New York, NY , USA: Association for Computing Machinery, 2019, p. 1955–1970. [Online]. Available: https://doi.org/10.1145/3319535.3354239

    work page doi:10.1145/3319535.3354239 2019

  9. [10]

    Security opera- tions center: A systematic study and open challenges,

    M. Vielberth, F. Böhm, I. Fichtinger, and G. Pernul, “Security opera- tions center: A systematic study and open challenges,”IEEE Access, vol. 8, pp. 227 756–227 779, 2020

    2020

  10. [11]

    Cost of a data breach report 2025,

    Ponemon Institute, “Cost of a data breach report 2025,” https://www. ibm.com/reports/data-breach, 2025, accessed: 2026-02-01

    2025

  11. [12]

    M-Trends 2025: Report,

    Mandiant, “M-Trends 2025: Report,”Mandiant Technical Report, 2025, accessed: 2026-01-25. 30

    2025

  12. [13]

    A survey on large language model based autonomous agents,

    L. Wang, C. Ma, X. Feng, Z. Zhang, H. Yang, J. Zhang, Z. Chen, J. Tang, X. Chen, Y . Lin, W. X. Zhao, Z. Wei, and J.-R. Wen, “A survey on large language model based autonomous agents,”Frontiers of Computer Science, vol. 18, no. 6, p. 186345, 2024

    2024

  13. [14]

    The Rise and Potential of Large Language Model Based Agents: A Survey

    Z. Xi, W. Chen, X. Guo, W. He, Y . Ding, B. Hong, M. Zhang, J. Wang, S. Jin, E. Zhouet al., “The rise and potential of large language model based agents: A survey,”arXiv preprint arXiv:2309.07864, 2023

    work page internal anchor Pith review arXiv 2023

  14. [15]

    and Nadiri, A

    Y . Talebirad and A. Nadiri, “Multi-agent collaboration: Harnessing the power of intelligent LLM agents,”arXiv preprint arXiv:2306.03314, 2023

    work page arXiv 2023

  15. [16]

    Yuanning Feng, Sinan Wang, Zhengxiang Cheng, Yao Wan, and Dongping Chen

    Z. Durante, Q. Huang, N. Wake, R. Gong, J. S. Cho, Y . Nishi, N. Kanezaki, and J. Gao, “Agent AI: Surveying the horizons of multimodal interaction,”arXiv preprint arXiv:2401.03568, 2024

    work page arXiv 2024

  16. [17]

    Log parsing with prompt-based few-shot learning,

    V .-H. Le and H. Zhang, “Log parsing with prompt-based few-shot learning,” inProceedings of the 45th International Conference on Software Engineering (ICSE). IEEE, 2023, pp. 2438–2449

    2023

  17. [18]

    LILAC: Log parsing using LLMs with adaptive parsing cache,

    Z. Jiang, J. Liu, Z. Chen, Y . Li, J. Huang, Y . Huo, P. He, J. Gu, and M. R. Lyu, “LILAC: Log parsing using LLMs with adaptive parsing cache,”Proceedings of the ACM on Software Engineering (FSE), vol. 1, pp. 1–24, 2024

    2024

  18. [19]

    arXiv preprint arXiv:2402.00891 (2024), https://arxiv.org/pdf/2402.00891

    F. N. Motlagh, M. Hajizadeh, M. Majd, P. Najafi, F. Cheng, and C. Meinel, “Large language models in cybersecurity: State-of-the-art,” arXiv preprint arXiv:2402.00891, 2024

    work page arXiv 2024

  19. [20]

    Li- nids: Llm-based intelligent nids rules generation for cybersecurity applications,

    A. Abdennebi, K. Nadjia, L. Lahlou, and H. Ould-Slimane, “Li- nids: Llm-based intelligent nids rules generation for cybersecurity applications,” in2025 16th International Conference on Network of the Future (NoF), 2025, pp. 55–63

    2025

  20. [21]

    SecureBERT: A domain-specific language model for cybersecurity,

    E. Aghaei, X. Niu, W. Shadid, and E. Al-Shaer, “SecureBERT: A domain-specific language model for cybersecurity,”Journal of Cyber- security and Privacy, vol. 3, no. 3, pp. 391–414, 2023

    2023

  21. [22]

    CyBERT: Contextualized embeddings for the cybersecurity domain,

    P. Ranade, A. Piplai, S. Mittal, A. Joshi, and T. Finin, “CyBERT: Contextualized embeddings for the cybersecurity domain,” in2021 IEEE International Conference on Big Data (Big Data). IEEE, 2021, pp. 3334–3342

    2021

  22. [23]

    SecureFalcon: The next cyber reasoning system for cyber security,

    M. A. Ferrag, N. Ndhlovu, N. Tihanyi, L. C. Cordeiro, M. Debbah, and T. Lestable, “SecureFalcon: The next cyber reasoning system for cyber security,”arXiv preprint arXiv:2307.06616, 2023

    work page arXiv 2023

  23. [24]

    Evaluating Large Language Models Trained on Code

    M. Chen, J. Tworek, H. Jun, Q. Yuan, H. P. d. O. Pinto, J. Kaplanet al., “Evaluating large language models trained on code,”arXiv preprint arXiv:2107.03374, 2021

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  24. [25]

    Code Llama: Open Foundation Models for Code

    B. Rozière, J. Gehring, F. Gloeckle, S. Sootla, I. Gat, X. E. Tan et al., “Code Llama: Open foundation models for code,”arXiv preprint arXiv:2308.12950, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  25. [26]

    Not what you’ve signed up for: Compromising real- world llm-integrated applications with indirect prompt injection,

    K. Greshake, S. Abdelnabi, S. Mishra, C. Endres, T. Holz, and M. Fritz, “Not what you’ve signed up for: Compromising real-world llm- integrated applications with indirect prompt injection,” ser. AISec ’23. New York, NY , USA: Association for Computing Machinery, 2023, p. 79–90. [Online]. Available: https://doi.org/10.1145/3605764.3623985

    work page doi:10.1145/3605764.3623985 2023

  26. [27]

    Prompt injection attack against LLM-integrated applications,

    Y . Liu, G. Deng, Z. Xu, Y . Li, Y . Zheng, Y . Zhang, L. Zhao, T. Zhang, K. Wang, and Y . Liu, “Prompt injection attack against LLM-integrated applications,” inProceedings of the 2024 USENIX Security Symposium, 2024, pp. 1–18

    2024

  27. [28]

    Ignore this title and HackAPrompt: Exposing systemic weaknesses of LLMs through a global scale prompt hacking competition,

    S. Schulhoff, J. Pinto, A. Khan, L.-F. Bouchard, C. Si, S. Anati, V . Tagliabue, A. Kost, C. Carnahan, and J. Boyd-Graber, “Ignore this title and HackAPrompt: Exposing systemic weaknesses of LLMs through a global scale prompt hacking competition,” 2023. [Online]. Available: https://aclanthology.org/2023.emnlp-main.302/

    2023

  28. [29]

    Beyond the safeguards: Exploring the security risks of ChatGPT,

    E. Derner and K. Batisti ˇc, “Beyond the safeguards: Exploring the security risks of ChatGPT,”arXiv preprint arXiv:2305.08005, 2023

    work page arXiv 2023

  29. [30]

    Owasp top 10 for llm apps & gen ai agentic security initiative,

    S. John, R. R. F. Del, K. Evgeniy, O. Helen, H. Idan, U. Kayla, H. Ken, S. Peter, A. Rakshith, B. Ronet al., “Owasp top 10 for llm apps & gen ai agentic security initiative,” Ph.D. dissertation, OW ASP, 2025

    2025

  30. [31]

    LLM agents can autonomously hack websites, 2024

    R. Fang, R. Bindu, A. Gupta, Q. Zhan, and D. Kang, “LLM agents can autonomously hack websites,” inarXiv preprint arXiv:2402.06664, 2024

    work page arXiv 2024

  31. [32]

    and Li, Z., 2024

    J. Xu, J. Nong, E. Vidal, A. Mukherjee, and X. Luo, “AutoAttacker: A large language model guided system to implement automatic cyber- attacks,”arXiv preprint arXiv:2403.01038, 2024

    work page arXiv 2024

  32. [33]

    A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions , volume=

    L. Huang, W. Yu, W. Ma, W. Zhong, Z. Feng, H. Wang, Q. Chen, W. Peng, X. Feng, B. Qin, and T. Liu, “A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,”ACM Trans. Inf. Syst., vol. 43, no. 2, Jan. 2025. [Online]. Available: https://doi.org/10.1145/3703155

    work page doi:10.1145/3703155 2025

  33. [34]

    Regulation 2024/1689 of the eur. parl. council of june 13, 2024 (eu artificial intelligence act),

    N. A. Smuha, “Regulation 2024/1689 of the eur. parl. council of june 13, 2024 (eu artificial intelligence act),”International Legal Materials, vol. 64, no. 5, p. 1234–1381, 2025

    2024

  34. [35]

    Artifi- cial intelligence risk management framework (AI RMF 1.0),

    National Institute of Standards and Technology, “Artifi- cial intelligence risk management framework (AI RMF 1.0),” https://nvlpubs.nist.gov/nistpubs/ai/NIST.AI.100-1. pdf?isid=enterprisehub_us&ikw=enterprisehub_us_lead% 2Fhow-to-responsibly-use-ai-powered-hr-tools_textlink_https%3A% 2F%2Fnvlpubs.nist.gov%2Fnistpubs%2Fai%2FNIST.AI.100-1.pdf, 2023, nIST AI 100-1

    2023

  35. [36]

    The operational role of security information and event management systems,

    S. Bhatt, P. K. Manadhata, and L. Zomlot, “The operational role of security information and event management systems,”IEEE Security & Privacy, vol. 12, no. 5, pp. 35–41, 2014

    2014

  36. [37]

    LangGraph: Building stateful multi-agent appli- cations,

    LangChain, Inc., “LangGraph: Building stateful multi-agent appli- cations,” https://github.com/langchain-ai/langgraph, 2024, accessed: 2025-10-15

    2024

  37. [38]

    Computer security incident handling guide (SP 800-61 rev. 3),

    National Institute of Standards and Technology, “Computer security incident handling guide (SP 800-61 rev. 3),” https://csrc.nist.gov/pubs/ sp/800/61/r3/final, 2024, accessed: 2026-03-01

    2024

  38. [39]

    Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone

    M. Abdin, J. Aneja, H. Awadalla, A. Awadallah, A. A. Awan, N. Bach et al., “Phi-3 technical report: A highly capable language model locally on your phone,”arXiv preprint arXiv:2404.14219, 2024

    work page internal anchor Pith review arXiv 2024

  39. [40]

    Mistral 7B

    A. Q. Jiang, A. Sablayrolles, A. Mensch, C. Bamford, D. S. Chaplot, D. de las Casas, F. Bressand, G. Lengyel, G. Lample, L. Saulnier, L. R. Lavaud, M.-A. Lachaux, P. Stock, T. L. Scao, T. Lavril, T. Wang, T. Lacroix, and W. E. Sayed, “Mistral 7b,” 2023. [Online]. Available: https://arxiv.org/abs/2310.06825

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  40. [41]

    Qlora: Efficient finetuning of quantized llms,

    T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, “Qlora: Efficient finetuning of quantized llms,” inAdvances in Neural Information Processing Systems, A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine, Eds., vol. 36. Curran Associates, Inc., 2023, pp. 10 088–10 115. [Online]. Available: https://proceedings.neurips.cc/paper_files...

    2023

  41. [42]

    LoRA: Low-rank adaptation of large language models,

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, and W. Chen, “LoRA: Low-rank adaptation of large language models,” in International Conference on Learning Representations (ICLR), 2022

    2022

  42. [43]

    Emerging threats open ruleset,

    Proofpoint, “Emerging threats open ruleset,” https://rules. emergingthreats.net/, 2024

    2024

  43. [44]

    Fast unfolding of communities in large networks,

    V . D. Blondel, J.-L. Guillaume, R. Lambiotte, and E. Lefebvre, “Fast unfolding of communities in large networks,”Journal of Statistical Mechanics: Theory and Experiment, vol. 2008, no. 10, p. P10008, 2008

    2008

  44. [45]

    Intelligence- driven computer network defense informed by analysis of adversary campaigns and intrusion kill chains,

    E. M. Hutchins, M. J. Cloppert, and R. M. Amin, “Intelligence- driven computer network defense informed by analysis of adversary campaigns and intrusion kill chains,”Leading Issues in Information Warfare & Security Research, vol. 1, no. 1, pp. 80–106, 2011

    2011

  45. [46]

    Constructing attack scenarios through correlation of intrusion alerts,

    P. Ning, Y . Cui, and D. S. Reeves, “Constructing attack scenarios through correlation of intrusion alerts,” inProceedings of the 9th ACM Conference on Computer and Communications Security, 2002, pp. 245–254

    2002

  46. [47]

    A systematic survey on multi- step attack detection,

    J. Navarro, A. Deruyver, and P. Parrend, “A systematic survey on multi- step attack detection,”Computers & Security, vol. 76, pp. 214–249, 2018

    2018

  47. [48]

    Model context protocol specification,

    Anthropic, “Model context protocol specification,” https: //modelcontextprotocol.io/, 2025, accessed: 2025-11-01

    2025

  48. [49]

    Ollama – get up and running with large language models locally,

    Ollama, “Ollama – get up and running with large language models locally,” https://ollama.com/, 2024

    2024

  49. [50]

    Streamlit — the fastest way to build and share data apps,

    Snowflake Inc., “Streamlit — the fastest way to build and share data apps,” https://streamlit.io/, 2024

    2024

  50. [51]

    A multi-vocal review of security orchestration,

    C. Islam, M. A. Babar, and S. Nepal, “A multi-vocal review of security orchestration,”ACM Computing Surveys, vol. 52, no. 2, pp. 1–45, 2019

    2019

  51. [52]

    Could SOAR save the SOC?

    R. Brewer, “Could SOAR save the SOC?”Computer Fraud & Security, vol. 2019, no. 9, pp. 8–11, 2019

    2019

  52. [53]

    Splunk SOAR documentation,

    Splunk (Cisco), “Splunk SOAR documentation,” https://docs.splunk. com/Documentation/SOAR/current/, 2024, accessed: 2026-03-01

    2024

  53. [54]

    Cortex XSOAR: Security orchestration, au- tomation and response,

    Palo Alto Networks, “Cortex XSOAR: Security orchestration, au- tomation and response,” https://docs-cortex.paloaltonetworks.com/r/ Cortex-XSOAR, 2024, accessed: 2026-03-01

    2024

  54. [55]

    Shuffle: Open source SOAR platform,

    Shuffle, “Shuffle: Open source SOAR platform,” https://shuffler.io/docs, 2024, accessed: 2026-03-01

    2024

  55. [56]

    CrowdStrike Falcon platform: Endpoint security and XDR,

    CrowdStrike, “CrowdStrike Falcon platform: Endpoint security and XDR,” https://www.crowdstrike.com/platform/, 2024, accessed: 2026- 03-01

    2024

  56. [57]

    Microsoft Sentinel: Cloud-native SIEM and SOAR,

    Microsoft, “Microsoft Sentinel: Cloud-native SIEM and SOAR,” https: //learn.microsoft.com/en-us/azure/sentinel/overview, 2024, accessed: 2026-03-01

    2024

  57. [58]

    Cortex XSIAM: AI-driven security operations platform,

    Palo Alto Networks, “Cortex XSIAM: AI-driven security operations platform,” https://www.paloaltonetworks.com/cortex/cortex-xsiam, 2024, accessed: 2026-03-01. 31

    2024

  58. [59]

    Physical Safety

    H. Xu, Y . Shi, Z. Zhao, Y . Wang, C. Xiao, B. Xu, J. Cheng, J. Gao, Q. Lin, F. Yang, Y . Qiao, and D. Lu, “Large language models for cyber security: A systematic literature review,”arXiv preprint arXiv:2405.04760, 2024

    work page arXiv 2024

  59. [60]

    Revolutionizing cyber threat detection with large language models: A privacy-preserving BERT-based lightweight model for IoT/IIoT devices,

    M. A. Ferrag, O. Friha, D. Hamouda, L. Maglaras, and H. Janicke, “Revolutionizing cyber threat detection with large language models: A privacy-preserving BERT-based lightweight model for IoT/IIoT devices,”IEEE Access, vol. 12, pp. 23 733–23 750, 2024

    2024

  60. [61]

    Getting PWNed by AI: Penetration testing with large language models,

    A. Happe and J. Cito, “Getting PWNed by AI: Penetration testing with large language models,” inProceedings of the 31st ACM Joint European Software Engineering Conference and Symposium on the Foundations of Software Engineering (ESEC/FSE). ACM, 2023, pp. 2082–2086

    2023

  61. [62]

    CyberMetric: A benchmark dataset for evaluating large language models knowledge in cybersecurity,

    N. Tihanyi, M. A. Ferrag, R. Jain, T. Bisztray, and M. Debbah, “CyberMetric: A benchmark dataset for evaluating large language models knowledge in cybersecurity,”arXiv preprint arXiv:2402.07688, 2024

    work page arXiv 2024

  62. [63]

    Sec-Llama: A com- pact fine-tuned LLM for network intrusion detection in Kubernetes clusters,

    A. Abdennebi, N. Kara, and H. Ould-Slimane, “Sec-Llama: A com- pact fine-tuned LLM for network intrusion detection in Kubernetes clusters,” in2025 IEEE International Mediterranean Conference on Communications and Networking (MeditCom). IEEE, 2025, pp. 1–6

    2025

  63. [64]

    Machine learning and large language models-based techniques for cyber threat detection: A comparative study,

    ——, “Machine learning and large language models-based techniques for cyber threat detection: A comparative study,” in2025 IEEE Inter- national Conference on Cyber Security and Resilience (CSR). IEEE, 2025, pp. 1–8

    2025

  64. [65]

    Deep learning for cyber security intrusion detection: Approaches, datasets, and comparative study,

    M. A. Ferrag, L. Maglaras, S. Moschoyiannis, and H. Janicke, “Deep learning for cyber security intrusion detection: Approaches, datasets, and comparative study,”Journal of Information Security and Applica- tions, vol. 50, p. 102419, 2020

    2020

  65. [66]

    Network intrusion detection system: A systematic study of machine learning and deep learning approaches,

    Z. Ahmad, A. Shahid Khan, C. Wai Shiang, J. Abdullah, and F. Ahmad, “Network intrusion detection system: A systematic study of machine learning and deep learning approaches,”Transactions on Emerging Telecommunications Technologies, vol. 32, no. 1, p. e4150, 2021

    2021

  66. [67]

    Kitsune: An ensemble of autoencoders for online network intrusion detection,

    Y . Mirsky, T. Doitshman, Y . Elovici, and A. Shabtai, “Kitsune: An ensemble of autoencoders for online network intrusion detection,” arXiv preprint arXiv:1802.09089, 2018

    work page arXiv 2018

  67. [68]

    Machine learning and deep learning methods for intrusion detection systems: A survey,

    H. Liu and B. Lang, “Machine learning and deep learning methods for intrusion detection systems: A survey,”Applied Sciences, vol. 9, no. 20, p. 4396, 2019

    2019

  68. [69]

    A hybrid intrusion detection system based on sparse autoencoder and deep neural network,

    K. Narayana Rao, K. Venkata Rao, and P. R. P.V .G.D., “A hybrid intrusion detection system based on sparse autoencoder and deep neural network,”Computer Communications, vol. 180, pp. 77– 88, 2021. [Online]. Available: https://www.sciencedirect.com/science/ article/pii/S0140366421003285

    2021

  69. [70]

    UNSW-NB15: A comprehensive data set for network intrusion detection systems (UNSW-NB15 network data set),

    N. Moustafa and J. Slay, “UNSW-NB15: A comprehensive data set for network intrusion detection systems (UNSW-NB15 network data set),” in2015 Military Communications and Information Systems Conference (MilCIS). IEEE, 2015, pp. 1–6

    2015

  70. [71]

    Toward generating a new intrusion detection dataset and intrusion traffic characterization,

    I. Sharafaldin, A. H. Lashkari, and A. A. Ghorbani, “Toward generating a new intrusion detection dataset and intrusion traffic characterization,” Proceedings of the 4th International Conference on Information Sys- tems Security and Privacy (ICISSP), pp. 108–116, 2018

    2018

  71. [72]

    Developing realistic distributed denial of service (DDoS) attack dataset and taxonomy,

    I. Sharafaldin, A. Habibi Lashkari, S. Hakak, and A. A. Ghorbani, “Developing realistic distributed denial of service (DDoS) attack dataset and taxonomy,” in2019 International Carnahan Conference on Security Technology (ICCST). IEEE, 2019, pp. 1–8

    2019

  72. [73]

    Generative deep learning to detect cyberattacks for the IoT-23 dataset,

    N. Abdalgawad, A. Sajun, Y . Kaddoura, I. A. Zualkernan, and F. Aloul, “Generative deep learning to detect cyberattacks for the IoT-23 dataset,”IEEE Access, vol. 10, pp. 6430–6441, 2022

    2022

  73. [74]

    Event log correlation for multi-step attack detection,

    S. U. Shaukat, S. Khan, and S. Parkinson, “Event log correlation for multi-step attack detection,”SECURITY AND PRIVACY, vol. 9, no. 1, p. e70151, 2026. [Online]. Available: https://onlinelibrary.wiley.com/ doi/abs/10.1002/spy2.70151

    work page doi:10.1002/spy2.70151 2026

  74. [75]

    Snort – network intrusion detection and prevention system,

    Cisco Talos, “Snort – network intrusion detection and prevention system,” https://www.snort.org/, 2024

    2024

  75. [76]

    Suricata – open source IDS/IPS/NSM engine,

    Open Information Security Foundation (OISF), “Suricata – open source IDS/IPS/NSM engine,” https://suricata.io/, 2024

    2024

  76. [77]

    Y ARA: The pattern matching swiss knife,

    V . Alvarez, “Y ARA: The pattern matching swiss knife,” https:// virustotal.github.io/yara/, 2024

    2024

  77. [78]

    Sigma: Generic signature format for SIEM systems,

    SigmaHQ, “Sigma: Generic signature format for SIEM systems,” https: //github.com/SigmaHQ/sigma, 2024, accessed: 2026-03-01

    2024

  78. [79]

    Event log correlation for multi-step attack detection,

    S. U. Shaukat, S. Khan, and S. Parkinson, “Event log correlation for multi-step attack detection,”Security and Privacy, vol. 9, no. 1, p. e70151, 2026

    2026

  79. [80]

    HOLMES: Real-time APT detection through correlation of suspicious information flows,

    S. M. Milajerdi, R. Gjomemo, B. Eshete, R. Sekar, and V . N. Venkatakr- ishnan, “HOLMES: Real-time APT detection through correlation of suspicious information flows,” in2019 IEEE Symposium on Security and Privacy (SP). IEEE, 2019, pp. 1137–1152

    2019

  80. [81]

    MITRE ATT&CK: Design and philosophy,

    B. E. Strom, A. Applebaum, D. P. Miller, K. C. Nickels, A. G. Pennington, and C. B. Thomas, “MITRE ATT&CK: Design and philosophy,”MITRE Technical Report, pp. 1–30, 2018

    2018

Showing first 80 references.