Hugging Face's logo

TextCortex
/
clef-cybersecurity

Text Classification
Transformers
Safetensors
English
German
cybersecurity
prompt-injection
data-exfiltration
clef
custom-code
Eval Results (legacy)
Model card Files
xet
 Community
clef-cybersecurity
File size: 3,561 Bytes
8f76a71
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
{
  "type": "aggregate_prompt_injection_benchmarks",
  "models": [
    {
      "model": "TextCortex/clef-cybersecurity",
      "name": "clef-cybersecurity",
      "full_en": {
        "n": 510,
        "attacks": 256,
        "clean": 254,
        "auroc": 0.9925335260826772,
        "caught": 232,
        "false_alarms": 0
      },
      "full_de": {
        "n": 510,
        "attacks": 256,
        "clean": 254,
        "auroc": 0.9743940698818898,
        "caught": 241,
        "false_alarms": 17
      },
      "skills_en": {
        "n": 48,
        "attacks": 27,
        "clean": 21,
        "auroc": 1.0,
        "caught": 23,
        "false_alarms": 0
      },
      "skills_de": {
        "n": 48,
        "attacks": 27,
        "clean": 21,
        "auroc": 0.9171075837742504,
        "caught": 26,
        "false_alarms": 3
      },
      "pdf": {
        "n": 730,
        "attacks": 107,
        "clean": 623,
        "auroc": 0.9856212778086137,
        "caught": 84,
        "false_alarms": 3
      },
      "decision_rule": "score > 0.5"
    },
    {
      "model": "jev",
      "name": "Jev",
      "full_en": {
        "auroc": 0.9799920029527559,
        "n": 510
      },
      "full_de": {
        "auroc": 0.9564391609251969,
        "n": 510
      },
      "skills_en": {
        "n": 48,
        "attacks": 27,
        "clean": 21,
        "auroc": 0.9841269841269841,
        "caught": 25,
        "false_alarms": 1
      },
      "skills_de": {
        "n": 48,
        "attacks": 27,
        "clean": 21,
        "auroc": 0.9603174603174603,
        "caught": 25,
        "false_alarms": 2
      },
      "pdf": {
        "n": 730,
        "attacks": 107,
        "clean": 623,
        "auroc": 0.978525674682348,
        "caught": 73,
        "false_alarms": 2
      },
      "decision_rule": "score > 0.5"
    },
    {
      "model": "laya-cybersec-r2a",
      "name": "Laya R2a",
      "full_en": {
        "n": 510,
        "attacks": 256,
        "clean": 254,
        "auroc": 0.9155004306102362,
        "caught": 140,
        "false_alarms": 2,
        "recall": 0.546875,
        "fpr": 0.007874015748031496
      },
      "full_de": {
        "n": 510,
        "attacks": 256,
        "clean": 254,
        "auroc": 0.8780065821850394,
        "caught": 151,
        "false_alarms": 9,
        "recall": 0.58984375,
        "fpr": 0.03543307086614173
      },
      "skills_en": {
        "n": 48,
        "attacks": 27,
        "clean": 21,
        "auroc": 0.927689594356261,
        "caught": 18,
        "false_alarms": 0
      },
      "skills_de": {
        "n": 48,
        "attacks": 27,
        "clean": 21,
        "auroc": 0.9329805996472663,
        "caught": 18,
        "false_alarms": 0
      },
      "pdf": {
        "n": 730,
        "attacks": 107,
        "clean": 623,
        "auroc": 0.8855927753859077,
        "caught": 81,
        "false_alarms": 0
      },
      "decision_rule": "score > 0.95"
    }
  ],
  "cohort": {
    "full_per_language": 510,
    "skills_per_language": 48,
    "pdf_attacks": 107,
    "clean_pdfs": 623
  },
  "scope": "Previously inspected internal regression sets; not a public leaderboard or fresh blind test.",
  "threshold_note": "Strict >0.5 for CLEF and Jev; strict >0.95 for Laya R2a. Recall and false alarms are not equal-threshold comparisons.",
  "selected_epoch": 3,
  "epochs_trained": 4,
  "clef_secondary_pdf_at_095": {
    "n": 730,
    "attacks": 107,
    "clean": 623,
    "auroc": 0.9856212778086137,
    "caught": 69,
    "false_alarms": 2
  }
}