| 1 | """ |
| 2 | Safari binary cookie extractor for macOS. |
| 3 | |
| 4 | Parses ~/Library/Cookies/Cookies.binarycookies (unencrypted binary format) |
| 5 | using only stdlib. Zero pip dependencies. |
| 6 | |
| 7 | Reference: github.com/mdegrazia/Safari-Binary-Cookie-Parser |
| 8 | """ |
| 9 | |
| 10 | from __future__ import annotations |
| 11 | |
| 12 | import io |
| 13 | import struct |
| 14 | import sys |
| 15 | from pathlib import Path |
| 16 | |
| 17 | # Mac epoch: 2001-01-01 00:00:00 UTC (not used for filtering, but documented) |
| 18 | _MAC_EPOCH_OFFSET = 978307200 # seconds between Unix epoch and Mac epoch |
| 19 | |
| 20 | _MAGIC = b"cook" |
| 21 | |
| 22 | |
| 23 | def _read_null_terminated(data: bytes, offset: int) -> str: |
| 24 | """Read a null-terminated string from data starting at offset.""" |
| 25 | end = data.find(b"\x00", offset) |
| 26 | if end == -1: |
| 27 | end = len(data) |
| 28 | return data[offset:end].decode("utf-8", errors="replace") |
| 29 | |
| 30 | |
| 31 | def _parse_cookie_record(data: bytes) -> dict | None: |
| 32 | """Parse a single cookie record. Returns dict with url, name, value, path or None.""" |
| 33 | if len(data) < 44: |
| 34 | return None |
| 35 | try: |
| 36 | (size,) = struct.unpack("<I", data[0:4]) |
| 37 | # flags at offset 4 (4 bytes, little-endian) — not needed for extraction |
| 38 | (url_offset,) = struct.unpack("<I", data[16:20]) |
| 39 | (name_offset,) = struct.unpack("<I", data[20:24]) |
| 40 | (path_offset,) = struct.unpack("<I", data[24:28]) |
| 41 | (value_offset,) = struct.unpack("<I", data[28:32]) |
| 42 | # expiry at offset 40 (8-byte double, little-endian) — not needed for filtering |
| 43 | # creation at offset 48 (8-byte double, little-endian) — not needed |
| 44 | |
| 45 | url = _read_null_terminated(data, url_offset) |
| 46 | name = _read_null_terminated(data, name_offset) |
| 47 | path = _read_null_terminated(data, path_offset) |
| 48 | value = _read_null_terminated(data, value_offset) |
| 49 | |
| 50 | return {"url": url, "name": name, "value": value, "path": path} |
| 51 | except (struct.error, IndexError, UnicodeDecodeError): |
| 52 | return None |
| 53 | |
| 54 | |
| 55 | def _parse_page(page_data: bytes) -> list[dict]: |
| 56 | """Parse a single page of cookies. Returns list of cookie dicts.""" |
| 57 | cookies = [] |
| 58 | if len(page_data) < 8: |
| 59 | return cookies |
| 60 | |
| 61 | # Page header: 4 bytes (always 00 00 01 00), then 4-byte LE cookie count |
| 62 | try: |
| 63 | (num_cookies,) = struct.unpack("<I", page_data[4:8]) |
| 64 | except struct.error: |
| 65 | return cookies |
| 66 | |
| 67 | # Sanity check |
| 68 | if num_cookies > 10000: |
| 69 | return cookies |
| 70 | |
| 71 | # Cookie offsets: array of 4-byte LE uint32 starting at offset 8 |
| 72 | offsets_end = 8 + num_cookies * 4 |
| 73 | if offsets_end > len(page_data): |
| 74 | return cookies |
| 75 | |
| 76 | for i in range(num_cookies): |
| 77 | off_start = 8 + i * 4 |
| 78 | try: |
| 79 | (cookie_offset,) = struct.unpack("<I", page_data[off_start : off_start + 4]) |
| 80 | except struct.error: |
| 81 | continue |
| 82 | |
| 83 | if cookie_offset >= len(page_data): |
| 84 | continue |
| 85 | |
| 86 | cookie_data = page_data[cookie_offset:] |
| 87 | record = _parse_cookie_record(cookie_data) |
| 88 | if record: |
| 89 | cookies.append(record) |
| 90 | |
| 91 | return cookies |
| 92 | |
| 93 | |
| 94 | def extract_safari_cookies_macos( |
| 95 | domain: str, cookie_names: list[str] |
| 96 | ) -> dict[str, str] | None: |
| 97 | """ |
| 98 | Extract cookies from Safari on macOS. |
| 99 | |
| 100 | Args: |
| 101 | domain: Registrable host to match, with or without a leading dot |
| 102 | (e.g. "x.com" or ".x.com"). A stored cookie host matches when it |
| 103 | equals the domain or is a subdomain of it; unrelated hosts that |
| 104 | merely contain the text (e.g. "x.com.evil.tld") do not. |
| 105 | cookie_names: List of cookie names to extract (e.g. ["auth_token", "ct0"]) |
| 106 | |
| 107 | Returns: |
| 108 | Dict mapping cookie name to value for found cookies, or None on failure. |
| 109 | """ |
| 110 | if sys.platform != "darwin": |
| 111 | return None |
| 112 | |
| 113 | cookie_paths = [ |
| 114 | Path.home() |
| 115 | / "Library" |
| 116 | / "Containers" |
| 117 | / "com.apple.Safari" |
| 118 | / "Data" |
| 119 | / "Library" |
| 120 | / "Cookies" |
| 121 | / "Cookies.binarycookies", |
| 122 | Path.home() / "Library" / "Cookies" / "Cookies.binarycookies", |
| 123 | ] |
| 124 | cookie_path = next((path for path in cookie_paths if path.exists()), cookie_paths[0]) |
| 125 | |
| 126 | try: |
| 127 | raw = cookie_path.read_bytes() |
| 128 | except FileNotFoundError: |
| 129 | return None |
| 130 | except PermissionError: |
| 131 | print( |
| 132 | "[safari] Permission denied reading Cookies.binarycookies. " |
| 133 | "Enable Full Disk Access for Terminal in System Settings > " |
| 134 | "Privacy & Security > Full Disk Access.", |
| 135 | file=sys.stderr, |
| 136 | ) |
| 137 | return None |
| 138 | except OSError: |
| 139 | return None |
| 140 | |
| 141 | return _parse_binary_cookies(raw, domain, cookie_names) |
| 142 | |
| 143 | |
| 144 | def _host_matches(stored_host: str, domain: str) -> bool: |
| 145 | """True when stored_host equals domain or is a subdomain of it. |
| 146 | |
| 147 | Safari stores domain cookies with a leading dot (".x.com") and host-only |
| 148 | cookies without one ("x.com"); both spellings are accepted on either side. |
| 149 | """ |
| 150 | host = stored_host.strip().lstrip(".").lower() |
| 151 | wanted = domain.strip().lstrip(".").lower() |
| 152 | if not host or not wanted: |
| 153 | return False |
| 154 | return host == wanted or host.endswith("." + wanted) |
| 155 | |
| 156 | |
| 157 | def _parse_binary_cookies( |
| 158 | raw: bytes, domain: str, cookie_names: list[str] |
| 159 | ) -> dict[str, str] | None: |
| 160 | """Parse raw binary cookie data. Separated for testability.""" |
| 161 | if len(raw) < 8: |
| 162 | return None |
| 163 | |
| 164 | # Validate magic |
| 165 | if raw[:4] != _MAGIC: |
| 166 | return None |
| 167 | |
| 168 | try: |
| 169 | (num_pages,) = struct.unpack(">I", raw[4:8]) |
| 170 | except struct.error: |
| 171 | return None |
| 172 | |
| 173 | if num_pages > 100000: |
| 174 | return None |
| 175 | |
| 176 | # Read page sizes (big-endian uint32 array) |
| 177 | page_sizes_end = 8 + num_pages * 4 |
| 178 | if page_sizes_end > len(raw): |
| 179 | return None |
| 180 | |
| 181 | page_sizes = [] |
| 182 | for i in range(num_pages): |
| 183 | off = 8 + i * 4 |
| 184 | try: |
| 185 | (ps,) = struct.unpack(">I", raw[off : off + 4]) |
| 186 | page_sizes.append(ps) |
| 187 | except struct.error: |
| 188 | return None |
| 189 | |
| 190 | # Parse each page |
| 191 | names_set = set(cookie_names) |
| 192 | result: dict[str, str] = {} |
| 193 | offset = page_sizes_end |
| 194 | |
| 195 | for ps in page_sizes: |
| 196 | if offset + ps > len(raw): |
| 197 | break |
| 198 | page_data = raw[offset : offset + ps] |
| 199 | cookies = _parse_page(page_data) |
| 200 | for c in cookies: |
| 201 | if _host_matches(c["url"], domain) and c["name"] in names_set: |
| 202 | result[c["name"]] = c["value"] |
| 203 | offset += ps |
| 204 | |
| 205 | if not result: |
| 206 | return None |
| 207 | |
| 208 | return result |
| 209 |