Create README.md

README.md

@@ -0,0 +1,53 @@
+---
+pretty_name: Benign and Malicious PowerShell Scripts
+task_categories:
+- text-classification
+tags:
+- powershell
+- cybersecurity
+- malware-detection
+- static-analysis
+- code-obfuscation
+- abstract-syntax-tree
+size_categories:
+- 10K<n<100K
+---
+
+# Benign and Malicious PowerShell Scripts
+
+## Dataset summary
+
+This dataset contains **30,595 high-quality PowerShell scripts** for binary benign/malicious classification and static-analysis research. It includes both unobfuscated and heavily obfuscated PowerShell, with obfuscation techniques including—but not limited to—multiple encoding schemes.
+
+All published samples were validated as parseable with Microsoft's official PowerShell parser, `System.Management.Automation.Language.Parser`, and can be converted into a PowerShell abstract syntax tree (AST) without executing the script.
+
+> **Safety warning:** This dataset contains potentially malicious PowerShell code. Treat every sample as untrusted. Do not execute samples on a host system. Use isolated, non-production environments and static-analysis tooling only.
+
+## 中文简介
+
+本数据集包含 **30,595 份高质量 PowerShell 脚本**，覆盖良性与恶意两类样本，适用于二分类、静态分析、恶意代码检测、混淆识别及基于 AST 的研究。
+
+数据中包含大量高质量混淆样本，混淆方式包括但不限于多种编码方式。所有公开样本均已验证能够通过微软官方 PowerShell 解析器 `System.Management.Automation.Language.Parser` 成功解析并提取抽象语法树（AST），无需执行脚本。
+
+> **安全提示：** 数据集中可能包含恶意代码。请将所有样本视为不可信内容，不要直接执行；建议仅在隔离、非生产环境中进行静态分析。
+
+## Key characteristics
+
+- **Task:** binary classification of PowerShell scripts
+- **Content:** benign and malicious `.ps1` files
+- **Total samples:** 30,595
+- **Obfuscation:** includes multiple encoding and other PowerShell obfuscation techniques
+- **Syntax quality:** every sample is parseable by the official PowerShell AST parser
+- **Representation:** one script per file, organized by numeric label
+- **Source split traceability:** original split, shard, and row identifiers are retained in filenames
+
+## Dataset structure
+
+The downloadable archive is `ast_ps1_by_label.tar.gz` (approximately 5.21 GB / 4.85 GiB compressed) and expands to:
+
+The dataset uses the following label convention:
+
+| Directory | Label | Samples |
+|---|---|---:|
+| `0/` | Benign | 15,173 |
+| `1/` | Malicious | 15,422 |