mirror of
https://github.com/mandiant/capa.git
synced 2026-02-05 03:16:56 -08:00
* web: index: add gif of capa running * index: add screencast of running capa produced via: ``` asciinema capa.cast ./capa Practical\ Malware\ Analysis\ Lab\ 01-01.dll_ <ctrl-d> agg --no-loop --theme solarized-light capa.cast capa.gif ``` * web: index: start to sketch out style * web: landing page * web: merge rules website * web: rules: update bootstrap and integrate rules * web: rules: use pygments to syntax highlight rules Use the Pygments syntax-highlighting library to parse and render the YAML rule content. This way we don't have to manually traverse the rule nodes and emit lists; instead, we rely on the fact that YAML is pretty easy for humans to read and let them consume it directly, with some text formatting to help hint at the types/structure. * web: rules: use capa to load rule content capa (the library) has routines for deserializing the YAML content into structured objects, which means we can use tools like mypy to find bugs. So, prefer to use those routines instead of parsing YAML ourselves. * web: rules: linters Run and fix the issues identified by the following linters: - isort - black - ruff - mypy * web: rules: add some links to rule page Add links to the following external resources: - GitHub rule source in capa-rules repo - VirusTotal search for matching samples * web: rules: accept ?q= parameter for initial search Update the rules landing page to accept a HTTP query parameter named "q" that specifies an initial search term to to pass to pagefind. This enables external pages link to rule searches. * web: rules: add link to namespace search * web: rules: use consistent header Import header from root capa landing page. * web: rules: add umami script * web: add initial whats new section, TODOs * web: rules: remove old images * changelog * CI: remove temporary branch push event triggers * Delete web/rules/public/css/bootstrap-4.5.2.min.css * Delete web/rules/public/js/bootstrap-4.5.2.min.js * Delete web/public/img/capa.cast * Rename readme.md to README.md * web: rules: add scripts to pre-commit configs * web: rules: add scripts to pre-commit configs * lints * ci: add temporary branch push trigger to get incremental builds * web: rules: assert start_dir must exist * ci: web: rules: deep checkout so we can get rule history * web: rules: check output of subprocess * web: rules: factor out common CSS * web: rules: fix header links * web: rules: only index rule content, not surrounding text * ci: web: remote temporary branch push trigger
344 lines
10 KiB
Python
344 lines
10 KiB
Python
"""
|
|
Copyright (C) 2024 Mandiant, Inc. All Rights Reserved.
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at: [package root]/LICENSE.txt
|
|
Unless required by applicable law or agreed to in writing, software distributed under the License
|
|
is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and limitations under the License.
|
|
"""
|
|
|
|
import sys
|
|
import random
|
|
import logging
|
|
from typing import Dict, List
|
|
from pathlib import Path
|
|
|
|
import capa.rules
|
|
|
|
logger = logging.getLogger(__name__)
|
|
|
|
start_dir = Path(sys.argv[1])
|
|
txt_file_path = Path(sys.argv[2])
|
|
out_dir = Path(sys.argv[3])
|
|
output_html_path = out_dir / "index.html"
|
|
|
|
assert start_dir.exists(), "input directory must exist"
|
|
assert txt_file_path.exists(), "file-modification txt file must exist"
|
|
assert out_dir.exists(), "output directory must exist"
|
|
|
|
predefined_colors = [
|
|
"#9CAFAA",
|
|
"#577590",
|
|
"#a98467",
|
|
"#D6DAC8",
|
|
"#adc178",
|
|
"#f4d35e",
|
|
"#85182a",
|
|
"#d6c399",
|
|
"#dde5b6",
|
|
"#8da9c4",
|
|
"#fcd5ce",
|
|
"#706993",
|
|
"#FBF3D5",
|
|
"#1a659e",
|
|
"#c71f37",
|
|
"#EFBC9B",
|
|
"#7e7f9a",
|
|
]
|
|
|
|
|
|
def read_file_paths(txt_file_path: Path):
|
|
categorized_files: Dict[str, List[Path]] = {
|
|
"modified in the last day": [],
|
|
"modified in the last week": [],
|
|
"modified in the last month": [],
|
|
"modified in the last three months": [],
|
|
"modified in the last year": [],
|
|
"older": [],
|
|
}
|
|
|
|
lines = txt_file_path.read_text(encoding="utf-8").splitlines()
|
|
|
|
current_category = None
|
|
for line in lines:
|
|
line = line.strip()
|
|
if not line:
|
|
continue
|
|
if "===" in line:
|
|
category = line.strip("=").strip()
|
|
if category in categorized_files:
|
|
current_category = category
|
|
else:
|
|
logger.warning("Unrecognized category '%s'", category)
|
|
current_category = None
|
|
elif current_category:
|
|
parts = line.split(" ", 1)
|
|
if len(parts) == 2:
|
|
file_path, last_modified_date_str = parts
|
|
categorized_files[current_category].append(Path(file_path))
|
|
else:
|
|
logger.warning("Skipping line due to unexpected format: %s", line)
|
|
|
|
return categorized_files
|
|
|
|
|
|
def parse_rule(file_path: Path):
|
|
rule = capa.rules.Rule.from_yaml_file(file_path)
|
|
|
|
return {
|
|
"name": rule.name,
|
|
"namespace": rule.meta.get("namespace", ""),
|
|
"authors": rule.meta.get("authors", []),
|
|
"path": file_path,
|
|
"filename": file_path.name,
|
|
}
|
|
|
|
|
|
def generate_color():
|
|
return "#{:06x}".format(random.randint(0, 0xFFFFFF))
|
|
|
|
|
|
def get_first_word(namespace):
|
|
return namespace.split("/")[0] if "/" in namespace else namespace
|
|
|
|
|
|
def generate_html(categories_data, color_map):
|
|
html_content = """<!DOCTYPE html>
|
|
<html lang="en">
|
|
<head>
|
|
<meta charset="UTF-8">
|
|
<meta name="viewport" content="width=device-width, initial-scale=1.0">
|
|
<title>capa rules</title>
|
|
<link rel="stylesheet" href="./pagefind/pagefind-ui.css">
|
|
<link rel="stylesheet" href="./pagefind/pagefind-modular-ui.css">
|
|
<link href="./pagefind/pagefind-ui.css" rel="stylesheet">
|
|
<link href="./css/poppins.css" rel="stylesheet">
|
|
<link href="./css/bootstrap-5.3.3.min.css" rel="stylesheet">
|
|
<link rel="icon" href="./img/favicon.png" type="image/x-icon"> <!-- Favicon -->
|
|
<script src="./pagefind/pagefind-ui.js"></script>
|
|
<script defer src="https://cloud.umami.is/script.js" data-website-id="0bb8ff9e-fbcc-4ee2-9f9f-b337a2e8cc7f"></script>
|
|
<link rel="stylesheet" type="text/css" href="./css/style.css">
|
|
<style>
|
|
body {
|
|
background-color: #ffffff;
|
|
font-family: 'Poppins', sans-serif;
|
|
margin: 0;
|
|
padding: 0;
|
|
}
|
|
|
|
.container-fluid {
|
|
padding: 0 40px;
|
|
}
|
|
|
|
.row {
|
|
display: grid;
|
|
grid-template-columns: repeat(4, 1fr);
|
|
gap: 10px;
|
|
}
|
|
|
|
.card-wrapper {
|
|
display: flex;
|
|
align-items: stretch;
|
|
}
|
|
|
|
.card {
|
|
background-color: #FFFFFF;
|
|
border-radius: 10px;
|
|
box-shadow: 0px 0px 0px 0px #E5EDFF;
|
|
padding: 10px;
|
|
transition: box-shadow 0.3s ease-in-out;
|
|
display: flex;
|
|
flex-direction: column;
|
|
width: 100%;
|
|
}
|
|
|
|
.card:hover {
|
|
box-shadow: 10px 10px 0px 0px #E5EDFF;
|
|
}
|
|
|
|
.thin-rectangle {
|
|
width: 15px;
|
|
height: 40px;
|
|
position: absolute;
|
|
top: 20px;
|
|
left: 0;
|
|
}
|
|
|
|
.namespace {
|
|
color: grey;
|
|
font-size: 13px;
|
|
}
|
|
|
|
.rule-name a {
|
|
color: black;
|
|
font-weight: bold;
|
|
font-size: 14.5px;
|
|
}
|
|
|
|
.rule-name a:hover {
|
|
text-decoration: none;
|
|
}
|
|
|
|
.authors {
|
|
color: black;
|
|
font-size: 13px;
|
|
}
|
|
|
|
.tags {
|
|
margin-top: 10px;
|
|
}
|
|
|
|
.tag {
|
|
background-color: hsl(210, 98%, 80%);
|
|
border: none;
|
|
border-radius: 20px;
|
|
color: black;
|
|
font: 600 1.05rem/1 "Poppins", sans-serif;
|
|
padding: 0.5em 1.5em;
|
|
margin-right: 5px;
|
|
display: inline-block;
|
|
font-size: 12px;
|
|
text-transform: lowercase;
|
|
transition: transform 0.3s;
|
|
}
|
|
|
|
.tag:hover {
|
|
transform: scale(1.1);
|
|
}
|
|
|
|
.card-body {
|
|
flex: 1;
|
|
}
|
|
|
|
a {
|
|
color: inherit;
|
|
text-decoration: none;
|
|
}
|
|
|
|
a:hover {
|
|
text-decoration: underline;
|
|
}
|
|
</style>
|
|
</head>
|
|
<body>
|
|
|
|
<header
|
|
class="d-flex flex-wrap justify-content-center py-1 mb-4 border-bottom fixed-top"
|
|
style="background-color: rgba(255,255,255,0.95);
|
|
box-shadow: 0 0.5rem 1rem rgba(0,0,0,0.05),inset 0 -1px 0 rgba(0,0,0,0.15);"
|
|
>
|
|
<a href="/capa/" class="d-flex align-items-center mb-3 mb-md-0 me-md-auto">
|
|
<img src="./img/logo.png" height=48 />
|
|
</a>
|
|
|
|
<ul class="nav nav-pills">
|
|
<li class="nav-item d-flex align-items-center"><a href="/capa/#rules" class="nav-link text-dark">Rules</a></li>
|
|
<li class="nav-item d-flex align-items-center"><a href="/capa/#examples" class="nav-link text-dark">Examples</a></li>
|
|
<li class="nav-item d-flex align-items-center"><a href="/capa/#download" class="nav-link text-dark">Download</a></li>
|
|
</ul>
|
|
</header>
|
|
|
|
<div class="container-fluid" style="margin-top: 5rem !important;">
|
|
<div id="search" class="my-4"></div>
|
|
"""
|
|
|
|
for category, files in categories_data.items():
|
|
if not files:
|
|
continue
|
|
|
|
html_content += f'<h4>{category}</h4><div class="row mb-4">'
|
|
cards_data = []
|
|
for file_path in files:
|
|
try:
|
|
card_data = parse_rule(file_path)
|
|
cards_data.append(card_data)
|
|
except Exception as e:
|
|
logger.error("error parsing %s: %s", file_path, e)
|
|
|
|
for card in cards_data:
|
|
first_word = get_first_word(card["namespace"])
|
|
rectangle_color = color_map[first_word]
|
|
file_name = card["filename"].rpartition(".yml")[0]
|
|
|
|
card_html = f"""
|
|
<div class="card-wrapper">
|
|
<div class="card">
|
|
<div class="thin-rectangle" style="background-color: {rectangle_color};"></div>
|
|
<div class="card-body">
|
|
<div class="namespace">{card['namespace']}</div>
|
|
<div class="rule-name"><a href="./{file_name}.html">{card['name']}</a></div>
|
|
<div class="authors">{', '.join(card['authors'])}</div>
|
|
</div>
|
|
</div>
|
|
</div>"""
|
|
|
|
html_content += card_html
|
|
|
|
num_cards = len(cards_data)
|
|
num_empty_cells = (4 - (num_cards % 4)) % 4
|
|
if num_empty_cells > 0:
|
|
for _ in range(num_empty_cells):
|
|
html_content += """
|
|
<div class="card-wrapper">
|
|
<div class="card" style="visibility: hidden;"></div>
|
|
</div>"""
|
|
|
|
html_content += "</div>"
|
|
|
|
html_content += """
|
|
</div>
|
|
|
|
<script>
|
|
window.addEventListener('DOMContentLoaded', (event) => {
|
|
const search = new PagefindUI({
|
|
element: "#search",
|
|
showSubResults: true,
|
|
showEmptyFilters: false,
|
|
excerptLength: 15,
|
|
});
|
|
|
|
const params = new URLSearchParams(window.location.search);
|
|
const q = params.get("q");
|
|
|
|
if (q) {
|
|
console.log("initial query:", q)
|
|
search.triggerSearch(q)
|
|
}
|
|
});
|
|
</script>
|
|
</body>
|
|
</html>"""
|
|
|
|
output_html_path.write_text(html_content, encoding="utf-8")
|
|
|
|
|
|
categories_data = read_file_paths(txt_file_path)
|
|
|
|
|
|
color_map = {}
|
|
used_colors = set(predefined_colors)
|
|
color_index = 0
|
|
|
|
|
|
all_files = [file for category in categories_data.values() for file in category]
|
|
for file_path in all_files:
|
|
try:
|
|
card_data = parse_rule(file_path)
|
|
first_word = get_first_word(card_data["namespace"])
|
|
if first_word not in color_map:
|
|
if color_index < len(predefined_colors):
|
|
color_map[first_word] = predefined_colors[color_index]
|
|
color_index += 1
|
|
else:
|
|
new_color = generate_color()
|
|
while new_color in used_colors:
|
|
new_color = generate_color()
|
|
color_map[first_word] = new_color
|
|
used_colors.add(new_color)
|
|
except Exception as e:
|
|
logger.error("error parsing %s: %s", file_path, e)
|
|
|
|
generate_html(categories_data, color_map)
|
|
logger.info("HTML file has been generated: %s", output_html_path)
|