| 1 |
--[[ |
| 2 |
HTML Fuzzy Hashing Helper Module |
| 3 |
|
| 4 |
This module provides helper functions for HTML fuzzy hash matching |
| 5 |
and phishing detection based on HTML structure vs. content mismatches. |
| 6 |
|
| 7 |
Use case: Detect phishing where HTML structure matches legitimate emails |
| 8 |
but CTA (Call-To-Action) domains are different. |
| 9 |
]] |
| 10 |
|
| 11 |
local exports = {} |
| 12 |
local lua_util = require "lua_util" |
| 13 |
|
| 14 |
--[[ |
| 15 |
Analyze fuzzy results to detect potential phishing based on: |
| 16 |
- Text content fuzzy match (high score) |
| 17 |
- HTML structure fuzzy match (high score) |
| 18 |
- But HTML CTA domains differ from known legitimate |
| 19 |
|
| 20 |
Returns: phishing_score, explanation |
| 21 |
]] |
| 22 |
exports.check_html_text_mismatch = function(task, fuzzy_results) |
| 23 |
local html_matches = {} |
| 24 |
local text_matches = {} |
| 25 |
|
| 26 |
-- Separate HTML and text fuzzy matches |
| 27 |
for _, res in ipairs(fuzzy_results or {}) do |
| 28 |
if res.type == 'html' then |
| 29 |
table.insert(html_matches, res) |
| 30 |
elseif res.type == 'txt' then |
| 31 |
table.insert(text_matches, res) |
| 32 |
end |
| 33 |
end |
| 34 |
|
| 35 |
-- Phishing scenario: high text match but low/no HTML match |
| 36 |
if #text_matches > 0 and #html_matches == 0 then |
| 37 |
local max_text_score = 0 |
| 38 |
for _, res in ipairs(text_matches) do |
| 39 |
if res.score > max_text_score then |
| 40 |
max_text_score = res.score |
| 41 |
end |
| 42 |
end |
| 43 |
|
| 44 |
-- High text match but no HTML match = suspicious |
| 45 |
if max_text_score > 0.7 then |
| 46 |
return max_text_score * 0.5, string.format( |
| 47 |
"Text fuzzy match (%.2f) without HTML match - possible CTA substitution", |
| 48 |
max_text_score) |
| 49 |
end |
| 50 |
end |
| 51 |
|
| 52 |
-- Inverse scenario: HTML match but no text match |
| 53 |
-- (Could be template with varying content - less suspicious) |
| 54 |
if #html_matches > 0 and #text_matches == 0 then |
| 55 |
local max_html_score = 0 |
| 56 |
for _, res in ipairs(html_matches) do |
| 57 |
if res.score > max_html_score then |
| 58 |
max_html_score = res.score |
| 59 |
end |
| 60 |
end |
| 61 |
|
| 62 |
-- This is expected for newsletters/notifications |
| 63 |
lua_util.debugm('fuzzy_html', task, |
| 64 |
'HTML match (%.2f) without text match - likely template variation', |
| 65 |
max_html_score) |
| 66 |
end |
| 67 |
|
| 68 |
return 0, nil |
| 69 |
end |
| 70 |
|
| 71 |
--[[ |
| 72 |
Check if message has suspicious HTML fuzzy pattern: |
| 73 |
- Known legitimate HTML structure |
| 74 |
- But text content is different or manipulated |
| 75 |
- Useful for brand protection |
| 76 |
|
| 77 |
Example: Amazon email template with phishing text |
| 78 |
]] |
| 79 |
exports.check_brand_hijack = function(task, html_fuzzy_result, text_fuzzy_result) |
| 80 |
if not html_fuzzy_result then |
| 81 |
return 0, nil |
| 82 |
end |
| 83 |
|
| 84 |
-- High HTML match = known template |
| 85 |
if html_fuzzy_result.score > 0.8 then |
| 86 |
-- Check if text is suspicious |
| 87 |
if not text_fuzzy_result or text_fuzzy_result.score < 0.3 then |
| 88 |
return html_fuzzy_result.score * 0.6, |
| 89 |
string.format("Known HTML template (%.2f) with unfamiliar text - possible brand hijacking", |
| 90 |
html_fuzzy_result.score) |
| 91 |
end |
| 92 |
end |
| 93 |
|
| 94 |
return 0, nil |
| 95 |
end |
| 96 |
|
| 97 |
return exports |