extract 0 0 0

AYI-NEDJIMI/edr-evasion-fr

published by @AYI-NEDJIMI

Alignment dataset tracked in the /datasets sub-catalog.

Lifetime downloads
319
Last 30 days
30
Likes
-
Size
101 KB
Created on HF
2026-02-13
Age
8 months ago

Description

Techniques d'Evasion EDR/XDR - Dataset Francais

Dataset bilingue complet sur les techniques d'evasion EDR/XDR, concu pour les professionnels red team et blue team. Ce dataset couvre les techniques offensives, les regles de detection, la comparaison de solutions EDR/XDR et des Q&A detailles.

Contenu du Dataset

1. techniques_evasion (~40 techniques)

Techniques d'evasion organisees par categorie :

Evasion Memoire : AMSI bypass, ETW patching, NTDLL… See the full description on the dataset page: https://huggingface.co/datasets/AYI-NEDJIMI/edr-evasion-fr.

Tags

task_categories:question-answeringtask_categories:text-classificationsource_datasets:originallanguage:frlicense:apache-2.0size_categories:n<1Kformat:parquetmodality:textlibrary:datasetslibrary:dasklibrary:polarslibrary:mlcroissantregion:uscybersecurityedr-evasionred-teamblue-teamdetection-engineeringayinedjimi-consultantsfr

README current version from Hugging Face


language:

  • fr
    license: apache-2.0
    tags:
  • cybersecurity
  • edr-evasion
  • red-team
  • blue-team
  • detection-engineering
  • ayinedjimi-consultants
  • fr
    pretty_name: EDR/XDR Evasion Techniques (Français)
    size_categories:
  • 1K<n<10K
    task_categories:
  • question-answering
  • text-classification
    authors:
  • name: Ayi NEDJIMI
    url: https://ayinedjimi-consultants.fr/bio.html
    source_datasets:
  • original

Techniques d'Evasion EDR/XDR - Dataset Francais

Dataset bilingue complet sur les techniques d'evasion EDR/XDR, concu pour les professionnels red team et blue team. Ce dataset couvre les techniques offensives, les regles de detection, la comparaison de solutions EDR/XDR et des Q&A detailles.

Contenu du Dataset

1. techniques_evasion (~40 techniques)

Techniques d'evasion organisees par categorie :

  • Evasion Memoire : AMSI bypass, ETW patching, NTDLL unhooking, injection shellcode, module stomping, Phantom DLL hollowing
  • Contournement des Hooks Userland : Direct syscalls (SysWhispers, Hell's Gate, Halo's Gate), manual mapping, hardware breakpoints, indirect syscalls
  • Sabotage de l'Agent EDR : PPL bypass, minifilter tampering, kernel callback removal, ETW provider disabling, WMI abuse
  • Living off the Land (LOLBins) : MSBuild, InstallUtil, Regsvr32, CMSTP, Mshta, certutil, bitsadmin, rundll32, PowerShell CLM bypass
  • Detection de Sandbox/VM : Hardware fingerprinting, timing-based detection, user interaction checks, environment artifacts
  • Evasion Reseau : Domain fronting, DNS tunneling, encrypted C2, legitimate service abuse, protocol impersonation

2. regles_detection (~25 regles)

Regles de detection multi-format :

  • Regles Sigma pour la detection generique
  • Requetes KQL pour Microsoft Defender for Endpoint
  • Regles YARA pour l'analyse memoire et statique
  • Configurations Sysmon pour la collecte de telemetrie

3. comparaison_edr (15 solutions)

Comparaison detaillee des principales solutions EDR/XDR du marche avec scores de resistance a l'evasion.

4. questions_reponses (80 Q&A)

Questions-reponses approfondies couvrant tous les aspects de l'evasion et de la detection EDR.

Utilisation

from datasets import load_dataset

dataset = load_dataset("AYI-NEDJIMI/edr-evasion-fr")

# Acceder aux techniques d'evasion
for technique in dataset["techniques_evasion"]:
    print(f"{technique['id']}: {technique['nom']}")

# Acceder aux regles de detection
for regle in dataset["regles_detection"]:
    print(f"{regle['type_regle']}: {regle['nom_detection']}")

# Q&A
for qa in dataset["questions_reponses"]:
    print(f"Q: {qa['question']}")
    print(f"R: {qa['answer'][:100]}...")

Articles de Reference

Ce dataset est base sur les recherches et articles publies par AYI-NEDJIMI Consultants :

A propos de l'Auteur

AYI-NEDJIMI Consultants est un cabinet de conseil en cybersecurite specialise dans le red teaming, le blue teaming, la detection engineering et la reponse aux incidents. Nos experts accompagnent les entreprises dans la securisation de leurs infrastructures contre les menaces avancees.

Licence

Ce dataset est distribue sous licence MIT. Utilisation a des fins educatives et de recherche en securite uniquement.

Citation

@dataset{ayinedjimi_edr_evasion_fr_2025,
  title={Techniques d'Evasion EDR/XDR - Dataset Francais},
  author={AYI-NEDJIMI Consultants},
  year={2025},
  publisher={HuggingFace},
  url={https://huggingface.co/datasets/AYI-NEDJIMI/edr-evasion-fr}
}

Author

Ayi NEDJIMI - Cybersecurity Consultant & Trainer | AI Expert

Related Articles

Free Cybersecurity Resources

Part of the Collection

This dataset is part of the Cybersecurity Datasets & Tools Collection by AYI-NEDJIMI Consultants.

README history 8 revisions

Every night we snapshot the README of every dataset in the catalog. When the SHA changes we archive the new version and diff it against the last. This is the evolving thought record of the alignment-data field: what the author decided to say about the corpus, and how that framing shifted over time.

  1. 2026-02-13Add deep backlinks and author section95907ba6.6 KB
    Loading...
  2. 2026-02-13Add schema.org SEO metadata6485e8f4.5 KB
    Loading...
  3. 2026-02-13Upload data/questions_reponses/train-00000-of-00001.parquet with huggingface_hubb25fbae4.5 KB
    Loading...
  4. 2026-02-13Upload data/comparaison_edr/train-00000-of-00001.parquet with huggingface_hube292d984.5 KB
    Loading...
  5. 2026-02-13Upload data/regles_detection/train-00000-of-00001.parquet with huggingface_hub60a7caa4.5 KB
    Loading...
  6. 2026-02-13Upload data/techniques_evasion/train-00000-of-00001.parquet with huggingface_hub2dd59f54.5 KB
    Loading...
  7. 2026-02-13Upload README.md with huggingface_hub539e1ce4.5 KB
    Loading...
  8. 2026-02-13initial commitdb1bc270 B
    Loading...

Used by abliterated models none yet

No abliterated models in our catalog have declared this dataset in their YAML metadata yet. This can mean the dataset is used but not declared, is used outside abliteration workflows, or is new. The link will populate automatically as models are indexed.

Related datasets same stage · extract

Read further

Metadata

License
apache-2.0

Repository

Download the dataset

0

Copy any of these snippets into your notebook or terminal. All three fetch directly from Hugging Face using your own credentials.

Recommended: datasets library (Python)
from datasets import load_dataset
ds = load_dataset("AYI-NEDJIMI/edr-evasion-fr")
Raw snapshot (Python)
from huggingface_hub import snapshot_download
snapshot_download(repo_id="AYI-NEDJIMI/edr-evasion-fr", repo_type="dataset")
git clone (requires git-lfs)
git clone https://huggingface.co/datasets/AYI-NEDJIMI/edr-evasion-fr
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration