Token Classification
Transformers
Safetensors
xlm-roberta
pii
privacy
redaction
anonymization
data-masking
named-entity-recognition
Instructions to use Tracer-ML/PIIX-560m with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Tracer-ML/PIIX-560m with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("token-classification", model="Tracer-ML/PIIX-560m")# pip install -U transformers accelerate # Load model directly from transformers import AutoTokenizer, AutoModelForTokenClassification tokenizer = AutoTokenizer.from_pretrained("Tracer-ML/PIIX-560m") model = AutoModelForTokenClassification.from_pretrained("Tracer-ML/PIIX-560m", device_map="auto") - Notebooks
- Google Colab
- Kaggle
File size: 8,232 Bytes
26d1584 | 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 | """Deterministic detectors for machine-format identifiers and secrets.
Written from public format definitions (RFCs, checksum algorithms, vendor token prefixes),
not from benchmark rows. Each rule returns typed character spans on the original text.
"""
import math
import re
TLD = (r'(?:com|org|net|edu|gov|mil|int|io|ai|co|dev|app|cloud|info|biz|xyz|online|site|tech|me|tv|cc|us|uk|de|fr|es|it|nl|pl|se|no|fi|dk|ch|at|be|cz|pt|gr|tr|ua|by|kz|ru|su|рф|cn|jp|kr|in|br|mx|ar|au|nz|ca|za|sg|hk|il|ae|eu|local|localdomain|internal|intranet|lan|corp|home|test|example|onion|svc|cluster)')
LABEL = r'[A-Za-z0-9](?:[A-Za-z0-9-]{0,61}[A-Za-z0-9])?'
HOST = rf'(?<![\w@.-])(?:{LABEL}\.)+{TLD}(?![\w-])(?!\.[A-Za-z0-9])'
FILE_EXT = {'txt', 'log', 'json', 'yaml', 'yml', 'xml', 'csv', 'py', 'js', 'ts', 'sh', 'md', 'conf', 'cfg', 'ini', 'env',
'pdf', 'doc', 'docx', 'xls', 'xlsx', 'png', 'jpg', 'gif', 'zip', 'gz', 'tar', 'sql', 'java', 'go', 'rb', 'php', 'html', 'css'}
EMAIL = re.compile(r'(?<![\w.+-])[\w.+-]+@(?:[\w-]+\.)+[\w-]{2,}(?![\w-])', re.U)
URL = re.compile(r'(?<![\w])(?:(?:https?|ftp|sftp|ssh|wss?|jdbc:[a-z]+|postgres(?:ql)?|mysql|mongodb(?:\+srv)?|redis|amqp)://|www\.)[^\s<>"\'`,;)\]}]+', re.I)
IPV4 = re.compile(r'(?<![\d.])(?:(?:25[0-5]|2[0-4]\d|1?\d?\d)\.){3}(?:25[0-5]|2[0-4]\d|1?\d?\d)(?:/\d{1,2}|:\d{2,5})?(?![\d.]*\d)')
IPV6 = re.compile(r'(?<![\w:])(?:[0-9A-Fa-f]{1,4}:){7}[0-9A-Fa-f]{1,4}(?![\w:])|(?<![\w:])(?:[0-9A-Fa-f]{1,4}:){1,6}:(?:[0-9A-Fa-f]{1,4}:){0,5}[0-9A-Fa-f]{1,4}(?![\w:])')
MAC = re.compile(r'(?<![\w:-])(?:[0-9A-Fa-f]{2}[:-]){5}[0-9A-Fa-f]{2}(?![\w:-])|(?<![\w.])(?:[0-9A-Fa-f]{4}\.){2}[0-9A-Fa-f]{4}(?![\w.])')
HOSTNAME = re.compile(HOST, re.U)
PHONE = re.compile(r'(?<![\w+])\+\d{1,3}[\s.-]?\(?\d{2,4}\)?(?:[\s.-]?\d{2,4}){2,4}(?!\w)|(?<![\w+(])\(\d{3,4}\)[\s.-]?\d{3}[\s.-]?\d{2,4}(?:[\s.-]?\d{2})?(?!\w)'
r'|(?<![\w.-])\d{3}[.-]\d{3}[.-]\d{4}(?![\w-])|(?<![\w.-])8[\s-]?\(?\d{3}\)?[\s-]?\d{3}[\s-]?\d{2}[\s-]?\d{2}(?![\w-])')
CARD = re.compile(r'(?<![\w-])(?:\d[ -]?){12,18}\d(?![\w-])')
IBAN = re.compile(r'(?<!\w)[A-Z]{2}\d{2}(?: ?[A-Z0-9]{4}){2,7}(?: ?[A-Z0-9]{1,4})?(?!\w)')
SNILS = re.compile(r'(?<![\w-])\d{3}-\d{3}-\d{3}[ -]\d{2}(?![\w-])')
INN = re.compile(r'(?i)(?<!\w)(?:инн|inn)\D{0,10}(\d{12}|\d{10})(?!\d)')
PASSPORT_RU = re.compile(r'(?i)(?<!\w)(?:паспорт\w*|passport)[^\n\d]{0,40}?(\d{2}\s?\d{2}\s?(?:№\s?)?\d{6})(?!\d)')
SSN = re.compile(r'(?<![\w-])(?!000|666|9\d\d)\d{3}-(?!00)\d{2}-(?!0000)\d{4}(?![\w-])')
PEM = re.compile(r'-----BEGIN [A-Z0-9 ]*(?:PRIVATE KEY|CERTIFICATE)[A-Z0-9 ]*-----.*?-----END [A-Z0-9 ]*(?:PRIVATE KEY|CERTIFICATE)[A-Z0-9 ]*-----', re.S)
JWT = re.compile(r'(?<![\w-])eyJ[\w-]{8,}\.eyJ[\w-]{8,}\.[\w-]{8,}(?![\w-])')
VENDOR_KEY = re.compile(
r'(?<![\w-])(?:A(?:KIA|SIA|GPA|IDA|ROA)[0-9A-Z]{16}|gh[pousr]_[A-Za-z0-9]{36,}|github_pat_[A-Za-z0-9_]{40,}|glpat-[\w-]{20,}'
r'|xox[abposr]-[A-Za-z0-9-]{10,}|sk-(?:ant-|proj-)?[A-Za-z0-9_-]{20,}|sk_(?:live|test)_[A-Za-z0-9]{16,}|pk_(?:live|test)_[A-Za-z0-9]{16,}'
r'|AIza[\w-]{35}|ya29\.[\w-]{20,}|SG\.[\w-]{16,}\.[\w-]{16,}|hf_[A-Za-z0-9]{30,}|npm_[A-Za-z0-9]{36}|dop_v1_[a-f0-9]{64}'
r'|\d{8,10}:AA[\w-]{33})(?![\w-])')
KEYVAL = re.compile(
r'(?i)(?<![A-Za-z])(?:pass(?:word|wd|phrase)?|pwd|secret(?:[_-]?key)?|api[_-]?key|access[_-]?key|auth[_-]?token|token|client[_-]?secret'
r'|private[_-]?key|credentials?|пароль|токен|секрет)\w*["\']?\s*(?:[:=]|=>|is|:=)\s*["\']?([^\s"\'<>,;]{6,})')
BEARER = re.compile(r'(?i)(?<!\w)(?:bearer|basic)\s+([A-Za-z0-9._~+/=-]{16,})')
# connection strings and URLs that embed a user name and password: the whole URL is a credential
CRED_URL = re.compile(r'(?<![\w])[a-z][a-z0-9+.-]{1,20}://[^\s:@/"\'<>]{1,64}:[^\s@/"\'<>]{1,128}@[^\s"\'<>`,;)\]}]+', re.I)
TOKEN = re.compile(r'(?<![\w/+=.-])[A-Za-z0-9+_=-]{28,}(?![\w/+=-])')
# Widely known public sites: a link to them identifies nobody, so URL and hostname rules skip them.
PUBLIC = re.compile(r'(?i)(?:^|[/.@])(?:github(?:usercontent)?\.com|gitlab\.com|bitbucket\.org|stackoverflow\.com|stackexchange\.com|google\.com|googleapis\.com|'
r'gstatic\.com|youtube\.com|microsoft\.com|windows\.net|azure\.com|amazonaws\.com|apple\.com|mozilla\.org|wikipedia\.org|npmjs\.(?:com|org)|'
r'pypi\.org|python\.org|nodejs\.org|docker\.(?:com|io)|kubernetes\.io|apache\.org|w3\.org|schema\.org|cloudflare\.com|jsdelivr\.net|unpkg\.com|'
r'readthedocs\.io|medium\.com|twitter\.com|x\.com|facebook\.com|linkedin\.com|reddit\.com|yandex\.ru|vk\.com|example\.(?:com|org|net)|'
r'localhost|shields\.io|travis-ci\.(?:org|com)|circleci\.com|sentry\.io|slack\.com|zoom\.us|atlassian\.(?:com|net)|oracle\.com|maven\.org|golang\.org|rust-lang\.org)(?=$|[/:?#])')
def public(value):
host = re.sub(r'^[a-z+.:-]*//', '', value, flags=re.I).split('/')[0].split('?')[0]
return bool(PUBLIC.search('/' + host.split('@')[-1].split(':')[0]))
def luhn(digits):
total = 0
for i, ch in enumerate(reversed(digits)):
d = int(ch)
if i % 2:
d = d * 2 - 9 if d > 4 else d * 2
total += d
return total % 10 == 0
def iban_ok(value):
s = value.replace(' ', '')
if not 15 <= len(s) <= 34:
return False
s = s[4:] + s[:4]
return int(''.join(str(int(c, 36)) for c in s)) % 97 == 1
def snils_ok(value):
d = re.sub(r'\D', '', value)
total = sum(int(c) * w for c, w in zip(d[:9], range(9, 0, -1)))
check = 0 if total % 101 in (100, 0) and total >= 100 else total % 101
if total < 100: check = total
elif total in (100, 101): check = 0
else: check = total % 101 % 100
return check == int(d[9:])
def entropy(s):
return -sum(s.count(c) / len(s) * math.log2(s.count(c) / len(s)) for c in set(s))
def secretish(s):
classes = sum(bool(re.search(p, s)) for p in (r'[a-z]', r'[A-Z]', r'\d'))
return classes == 3 and entropy(s) >= 4.0 and not re.fullmatch(r'[0-9a-fA-F-]{32,36}', s)
def find(text):
out = []
def add(m, label, group=0, ok=True):
if ok and m.end(group) > m.start(group):
out.append(dict(start=m.start(group), end=m.end(group), label=label))
for m in CRED_URL.finditer(text): add(m, 'secret_url')
for m in PEM.finditer(text): add(m, 'secret_pem')
for m in JWT.finditer(text): add(m, 'secret_jwt')
for m in VENDOR_KEY.finditer(text): add(m, 'secret_vendor')
for m in KEYVAL.finditer(text):
v = m.group(1)
add(m, 'secret_kv', 1, not re.fullmatch(r'(?i)(true|false|null|none|required|changeme|string|undefined|\*+|x+|<.*>|\$\{?\w+\}?|%\w+%|\{\{.*\}\}|process\.env.*|os\.environ.*)', v))
for m in BEARER.finditer(text): add(m, 'secret_bearer', 1)
for m in EMAIL.finditer(text): add(m, 'email')
urls = []
for m in URL.finditer(text):
end = m.end()
while end > m.start() and text[end - 1] in '.,:;!?': end -= 1
urls.append((m.start(), end))
if not public(text[m.start():end]): out.append(dict(start=m.start(), end=end, label='url'))
for m in TOKEN.finditer(text): # free-standing high-entropy strings only, never fragments of a URL
add(m, 'secret_token', 0, secretish(m.group(0)) and not any(a <= m.start() < b for a, b in urls))
for m in IPV4.finditer(text): add(m, 'ip')
for m in IPV6.finditer(text): add(m, 'ip', 0, m.group(0).count(':') >= 2)
for m in MAC.finditer(text): add(m, 'mac')
for m in HOSTNAME.finditer(text):
add(m, 'hostname', 0, m.group(0).rsplit('.', 1)[-1].lower() not in FILE_EXT and not public(m.group(0)) and not any(a <= m.start() < b for a, b in urls))
for m in PHONE.finditer(text):
add(m, 'phone', 0, 10 <= len(re.sub(r'\D', '', m.group(0))) <= 15)
for m in CARD.finditer(text):
d = re.sub(r'\D', '', m.group(0)); add(m, 'card_number', 0, 13 <= len(d) <= 19 and luhn(d))
for m in IBAN.finditer(text): add(m, 'iban', 0, iban_ok(m.group(0)))
for m in SNILS.finditer(text): add(m, 'snils')
for m in INN.finditer(text): add(m, 'inn', 1)
for m in PASSPORT_RU.finditer(text): add(m, 'passport', 1)
for m in SSN.finditer(text): add(m, 'ssn')
return out
|