tiangler commited on
Commit
857c7a4
·
verified ·
1 Parent(s): d1671da

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +117 -0
README.md ADDED
@@ -0,0 +1,117 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ annotations_creators:
3
+ - no-annotation
4
+ language:
5
+ - zh
6
+ language_creators:
7
+ - found
8
+ license:
9
+ - apache-2.0
10
+ multilinguality:
11
+ - monolingual
12
+ size_categories:
13
+ - 10K<n<100K
14
+ source_datasets:
15
+ - original
16
+ task_categories:
17
+ - text-classification
18
+ task_ids:
19
+ - text-classification
20
+ paperswithcode_id: null
21
+ pretty_name: Security Alert Classification Dataset
22
+ tags:
23
+ - security
24
+ - alert
25
+ - classification
26
+ - chinese
27
+ ---
28
+
29
+ # Dataset Card for Security Alert Classification Dataset
30
+
31
+ ## Dataset Description
32
+
33
+ - **Repository:** [N/A]
34
+ - **Paper:** [N/A]
35
+ - **Point of Contact:** [N/A]
36
+
37
+ ### Dataset Summary
38
+
39
+ 该数据集包含安全告警日志数据,用于训练大模型判断安全告警是真实攻击还是误报。数据集采用Alpaca格式,包含instruction、input和output三个字段。
40
+
41
+ ### Supported Tasks and Leaderboards
42
+
43
+ - **Task:** 安全告警分类
44
+ - **Task Type:** 文本分类
45
+ - **Languages:** 中文
46
+
47
+ ### Languages
48
+
49
+ 数据集中的文本为中文。
50
+
51
+ ## Dataset Structure
52
+
53
+ ### Data Instances
54
+
55
+ 每个样本包含以下字段:
56
+ - instruction: 任务说明,指导模型作为网络安全告警分析专家分析安全告警日志
57
+ - input: 告警日志数据(JSON格式),包含多种安全告警的详细信息
58
+ - output: 标签("攻击"或"误报")
59
+
60
+ ### Data Fields
61
+
62
+ - instruction: 字符串,任务说明
63
+ - input: 字符串,JSON格式的告警日志数据,包含告警来源、攻击类型、漏洞类型、危害等级、payload等信息
64
+ - output: 字符串,分类标签
65
+
66
+ ### Data Splits
67
+
68
+ - 训练集:65134条样本
69
+ - 攻击样本:291条 (0.45%)
70
+ - 误报样本:64843条 (99.55%)
71
+
72
+ ## Dataset Creation
73
+
74
+ ### Curation Rationale
75
+
76
+ 该数据集用于训练大模型进行安全告警分类,帮助安全分析师快速识别真实攻击和误报。
77
+
78
+ ### Source Data
79
+
80
+ #### Initial Data Collection and Normalization
81
+
82
+ 原始数据来自安全告警系统,包含各种类型的安全告警,如SQL注入、命令执行、信息泄露、扫描行为等。
83
+
84
+ #### Who are the source language producers?
85
+
86
+ 安全分析师
87
+
88
+ ### Annotations
89
+
90
+ #### Annotation process
91
+
92
+ 由安全分析师人工标注
93
+
94
+ #### Who are the annotators?
95
+
96
+ 安全分析师
97
+
98
+ ### Personal and Sensitive Information
99
+
100
+ 数据集中的IP地址和MAC地址等敏感信息已存在,但未进行进一步脱敏处理。
101
+
102
+ ## Considerations for Using the Data
103
+
104
+ ### Social Impact of Dataset
105
+
106
+ 该数据集可以帮助提高安全告警分析的效率,减少误报带来的资源浪费。
107
+
108
+ ### Discussion of Biases
109
+
110
+ 数据集存在极度类别不平衡问题,攻击样本仅占0.45%,误报样本占99.55%。
111
+
112
+ ### Other Known Limitations
113
+
114
+ 1. 数据集规模较大,但攻击样本极少,类别极度不平衡
115
+ 2. 安全告警类型可能不全面
116
+ 3. 数据集可能存在地区偏见,主要来自中国网络环境
117
+ 4. 标注可能存在一定的主观性