File size: 7,761 Bytes
e4e8a91
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
156
157
158
159
160
161
162
163
164
165
166
167
168
import re
from typing import List, Tuple

from rapidfuzz import fuzz

# 弱訊號詞: 單獨出現時常見於正常付款、通知或客服問答,不應直接視為高風險
MONEY_RULE = re.compile(
    r"\$|\bUSD\b|TWD|NT|錢|金額|台幣|美金|人民幣|日圓|匯款|匯入|佣金|手續費|未繳清|未繳費|"
    r"交易連結|交易鏈結|入.{0,3}帳|出.{0,3}金|付.{0,3}費|付.{0,3}款|付.{0,3}現|"
    r"繳.{0,3}費|刷.{0,3}卡|退.{0,3}款|退.{0,3}費|轉.{0,3}帳"
)
OTP_RULE = re.compile(r"OTP|驗證信|驗證碼|動態密碼|授權碼|一次性密碼")
ACCOUNT_RULE = re.compile(r"帳號|銀行|帳戶")

# 高風險情境詞: 與詐騙語境強相關
URGENCY_RULE = re.compile(r"立即|馬上|立刻|趕快|限時|最後|逾期|否則|將停用|凍結|封鎖|24小時|今日截止|搶購|盡速前往")
INVEST_RULE = re.compile(r"(投資|飆股|內線|代操|保證獲利|高報酬|穩賺|跟單|當沖|虛擬貨幣|以太幣|比特幣|BTC|USDT)")
IMPERSONATE_RULE = re.compile(r"(客服|官方|有誤|升級|操作不當|操作異常|帳戶異常|系統異常|登入異常|風控|安全驗證|司法|檢警|刑事|警察|法院|傳票|偵查|金管會)")
SENSITIVE_RULE = re.compile(r"(身分證|登入密碼|網銀密碼|交易密碼|信用卡|卡號|三碼|CVV|戶頭|提款卡|存摺|個資|證件)")
GIFT_RULE = re.compile(r"(免費|免費LINE貼圖|中獎|抽獎|贈品|領取|免費領取|點我領|兌換|禮物卡|點數|Steam卡)")
JOB_RULE = re.compile(r"(打工|兼職|在家工作|日領|週領|無經驗|免經驗|高薪|刷流水|刷單|打字員)")
TIME_RULE = re.compile(r"\d+秒|\d+分|^\d{2}:\d{2}(\d{2})?$")

# 反詐騙及正規金融內容的白名單/降權規則
DEBUNK_RULE = re.compile(r"反詐騙|165|手法|破解|多問|多查|手法分析|防範|防詐|提醒民眾|MyGoPen|cofacts|宣導")
APR_RULE = re.compile(r"年百分率(\s*\d{1,2}(\.\d{1,2})?%\s*\~\s*\d{1,2}(\.\d{1,2})?%)?|總費用年百分率(\s*\d{1,2}(\.\d{1,2})?%\s*\~\s*\d{1,2}(\.\d{1,2})?%)?|APR")
CACL_RULE = re.compile(r"試算|交易稅|機動計息|本息攤還|每月還款金額|每日還款金額|收入負債比")
DISCLAIMER_RULE = re.compile(r"免責聲明|主管機關|保有核貸與否權利|審慎評估還款能力|實際貸款額度及利率|參閱|產品公開說明書")
PERMISSION_RULE = re.compile(r"免(自行)?上傳(財力|證明)|應用金融科技|授權本行")
NORMAL_PAYMENT_RULE = re.compile(
    r"付款成功|付款完成|完成付款|繳費完成|扣款成功|刷卡成功|交易成功|訂單|訂購|發票|"
    r"收據|消費通知|入帳通知|轉帳成功|退款完成|付款通知|帳單|繳款通知"
)
OTP_SAFETY_RULE = re.compile(r"請勿|不要|切勿|勿將|勿提供|不要提供|勿告知|請勿告知|本人操作|非本人操作|驗證碼僅供本人")
QUESTION_RULE = re.compile(r"請問|想問|想請教|是否|怎麼|如何|可以|能否|嗎|呢|\?+$|?+$")
BANK_BRAND_RULE = re.compile(r"銀行|本行|信用卡|客服專線|客服中心")
BANK_TXN_NOTIFICATION_RULE = re.compile(
    r"線上交易|交易驗證碼|本次交易|商店名稱|交易內容|完成付款|有效時間|"
    r"交易頁面|客服專線|如非本人操作|請確認交易內容"
)

WEAK_SIGNAL_RULES: List[Tuple[str, re.Pattern, int, str]] = [
    ("money", MONEY_RULE, 12, "金流交易相關字詞"),
    ("otp", OTP_RULE, 12, "驗證碼相關字詞"),
    ("account", ACCOUNT_RULE, 10, "帳戶金融相關字詞"),
]

STRONG_SIGNAL_RULES: List[Tuple[str, re.Pattern, int, str]] = [
    ("urgency", URGENCY_RULE, 35, "催促或時效壓力用詞"),
    ("invest", INVEST_RULE, 45, "投資獲利相關誘導字詞"),
    ("impersonate", IMPERSONATE_RULE, 55, "假冒官方客服或異常通知用詞"),
    ("sensitive", SENSITIVE_RULE, 45, "敏感金融個資相關用詞"),
    ("gift", GIFT_RULE, 35, "贈品中獎相關誘導用詞"),
    ("job", JOB_RULE, 30, "高薪兼職相關誘導用詞"),
    ("time", TIME_RULE, 20, "時間壓力或操作時限用詞"),
]

BANK_RULES: List[Tuple[re.Pattern, str]] = [
    (APR_RULE, "年百分率、總費用年百分率等相關用詞"),
    (CACL_RULE, "試算款項等相關用詞"),
    (DISCLAIMER_RULE, "免責聲明等相關用詞"),
    (PERMISSION_RULE, "授權與免責等相關用詞"),
]

# 僅保留辨識力較高的模糊詞,避免「驗證碼」這類正常場景常見詞直接加權
FUZZY_KEYWORDS = [
    "保證獲利",
    "高報酬",
    "帳戶異常",
    "虛擬貨幣",
    "代操",
    "會員指定任務",
    "抽獎機會",
    "數量有限",
    "最後機會",
    "認證失敗",
    "點我領取",
    "限時搶購",
    "立即行動",
    "官方客服",
    "系統升級",
    "司法調查",
    "金管會公告",
]


def compare_rules(text: str) -> Tuple[int, str, List[str]]:
    is_debunk = bool(len(re.findall(DEBUNK_RULE, text)) >= 2)
    if is_debunk:
        return 20, ",含反詐騙宣導用詞", []

    reason_list: List[str] = []
    match = 0
    for pattern, reason in BANK_RULES:
        counter = len(re.findall(pattern, text))
        if counter >= 2:
            match += 1
            reason_list.append(reason)
        if match == 2:
            return 20, "含正規銀行所常用之" + "、".join(reason_list), []

    bank_notification_markers = 0
    for pattern in (BANK_BRAND_RULE, BANK_TXN_NOTIFICATION_RULE, OTP_SAFETY_RULE, NORMAL_PAYMENT_RULE):
        if re.search(pattern, text):
            bank_notification_markers += 1

    if bank_notification_markers >= 3 and re.search(MONEY_RULE, text) and re.search(OTP_RULE, text):
        return 20, "含正規銀行交易驗證通知語境", []

    score = 0
    signal_hits: List[str] = []
    reason_list = []

    has_payment_context = bool(re.search(NORMAL_PAYMENT_RULE, text))
    has_otp_safety_context = bool(re.search(OTP_SAFETY_RULE, text))
    has_question_context = text.endswith(("?", "?")) or bool(re.search(QUESTION_RULE, text))

    strong_signal_count = 0

    for kw in FUZZY_KEYWORDS:
        similarity = fuzz.partial_ratio(kw, text)
        if similarity > 85:
            score += 18
            signal_hits.append(f"fuzzy:{kw}")
            reason_list.append(kw)

    for key, pattern, point, sentence in STRONG_SIGNAL_RULES:
        if re.findall(pattern, text):
            score += point
            strong_signal_count += 1
            signal_hits.append(key)
            reason_list.append(sentence)

    for key, pattern, point, sentence in WEAK_SIGNAL_RULES:
        if not re.findall(pattern, text):
            continue

        adjusted_point = point
        if key == "money" and (has_payment_context or has_question_context):
            adjusted_point = 5
        elif key == "otp" and (has_otp_safety_context or has_question_context):
            adjusted_point = 4
        elif key == "account" and has_question_context:
            adjusted_point = 4

        score += adjusted_point
        signal_hits.append(key)
        reason_list.append(sentence)

    low_signal_combo = {"money", "otp", "account"}.intersection(signal_hits)
    if strong_signal_count >= 1 and low_signal_combo:
        score += 18
        reason_list.append("金流或驗證資訊與高風險情境交互出現")

    if strong_signal_count == 0 and low_signal_combo:
        if has_payment_context or has_otp_safety_context:
            score = min(score, 18)
        elif has_question_context:
            score = min(score, 20)
        else:
            score = min(score, 28)

    dedup_hits = list(dict.fromkeys(signal_hits))
    dedup_reasons = list(dict.fromkeys(reason_list))
    reason = "疑似包含" + "、".join(dedup_reasons) if dedup_reasons else ""

    return min(score, 100), reason, dedup_hits