Enhance profanity filtering to include hate symbol detection and censorship

- Updated the profanity filter to check for hate symbols (e.g., swastika Unicode) in addition to word-based profanity, ensuring comprehensive message filtering.
- Modified the `censor` function to replace hate symbols with `***`, maintaining functionality even when the `better-profanity` package is unavailable.
- Updated documentation in `discord-bridge.md` and `telegram-bridge.md` to reflect the new hate symbol handling features.
- Added tests to verify detection and censorship of hate symbols, ensuring robustness of the profanity filter.

docs/discord-bridge.md

@@ -86,7 +86,8 @@ Preview: `https://api.dicebear.com/7.x/{style}/png?seed=YourName`
   - **drop** (default): Do not bridge messages that contain profanity (message is dropped).
   - **censor**: Replace profanity with `****` and bridge the message.
   - **off**: No filtering; bridge all messages as-is.
-- Requires the `better-profanity` package (see `requirements.txt`). If the package is not installed and `filter_profanity` is `drop` or `censor`, a warning is logged and messages are bridged without filtering.
+- The filter checks word-based profanity (via `better-profanity` and optional `unidecode` for homoglyphs) and blocked hate symbols (e.g. swastika Unicode 卐/卍). Symbols are replaced with `***`.
+- Requires the `better-profanity` package (see `requirements.txt`). If the package is not installed and `filter_profanity` is `drop` or `censor`, a warning is logged and messages are bridged without word filtering; hate symbols are still filtered even without the package.
 
 ---
 

docs/telegram-bridge.md

@@ -76,7 +76,7 @@ Send a message on the bridged MeshCore channel — it should appear in the Teleg
 | `parse_mode` | No | `HTML` (default), `Markdown`, or `MarkdownV2` |
 | `disable_web_page_preview` | No | `true`/`false` — disable link previews (default: false) |
 | `max_message_length` | No | 1–4096; truncate longer messages (default: 4096) |
-| `filter_profanity` | No | Profanity handling: `drop` (default, do not bridge), `censor` (replace with ****), or `off`. Requires `better-profanity` package. |
+| `filter_profanity` | No | Profanity handling: `drop` (default), `censor`, or `off`. Word list via `better-profanity`; hate symbols (e.g. 卐/卍) are always blocked/censored. |
 
 \* Either `api_token` in config or `TELEGRAM_BOT_TOKEN` in the environment must be set when the bridge is enabled.
 

modules/profanity_filter.py

@@ -5,10 +5,18 @@ Shared profanity filter for bridge services (Discord, Telegram).
 Uses better-profanity when available; gracefully falls back to no-op if not installed.
 Uses unidecode when available to normalize Unicode (e.g. homoglyphs) to ASCII so
 better-profanity can detect them.
+Also checks for hate symbols (e.g. swastika Unicode) that word lists do not catch.
 """
 
 from typing import Optional
 
+# Unicode code points for symbols we treat as profanity (e.g. swastika forms).
+# These are checked in addition to better-profanity's word list.
+_HATE_SYMBOL_CODEPOINTS = frozenset({
+    0x5350,  # 卐 CJK swastika
+    0x534D,  # 卍 CJK swastika (reversed)
+})
+
 _profanity_available = False
 _profanity_initialized = False
 _warned_unavailable = False
@@ -27,6 +35,22 @@ except ImportError:
     unidecode = None  # type: ignore
 
 
+def _has_hate_symbols(text: str) -> bool:
+    """Return True if text contains any blocked hate-symbol code point."""
+    for cp in _HATE_SYMBOL_CODEPOINTS:
+        if chr(cp) in text:
+            return True
+    return False
+
+
+def _replace_hate_symbols(text: str, replacement: str = "***") -> str:
+    """Replace any hate-symbol code point in text with replacement."""
+    result = text
+    for cp in _HATE_SYMBOL_CODEPOINTS:
+        result = result.replace(chr(cp), replacement)
+    return result
+
+
 def _normalize_for_profanity(text: str) -> str:
     """Convert Unicode to ASCII when unidecode is available (catches homoglyph slurs)."""
     if _unidecode_available and unidecode is not None:
@@ -55,6 +79,7 @@ def _ensure_initialized(logger: Optional[object] = None) -> bool:
 def censor(text: Optional[str], logger: Optional[object] = None) -> str:
     """
     Replace profanity in text with ****. Returns original text if library unavailable.
+    Hate symbols (e.g. swastika Unicode) are replaced with ***.
 
     Args:
         text: Input string (message or username).
@@ -69,6 +94,8 @@ def censor(text: Optional[str], logger: Optional[object] = None) -> str:
         return str(text)
     if not text.strip():
         return text
+    # Replace hate symbols first (no dependency on better-profanity)
+    text = _replace_hate_symbols(text)
     if not _ensure_initialized(logger):
         return text
     normalized = _normalize_for_profanity(text)
@@ -77,17 +104,19 @@ def censor(text: Optional[str], logger: Optional[object] = None) -> str:
 
 def contains_profanity(text: Optional[str], logger: Optional[object] = None) -> bool:
     """
-    Return True if text contains any word from the profanity wordlist.
+    Return True if text contains any word from the profanity wordlist or a blocked hate symbol.
 
     Args:
         text: Input string to check.
         logger: Optional logger for one-time warning when better_profanity is not installed.
 
     Returns:
-        True if profanity detected, False otherwise or if library unavailable.
+        True if profanity or hate symbol detected, False otherwise or if library unavailable.
     """
     if text is None or not isinstance(text, str) or not text.strip():
         return False
+    if _has_hate_symbols(text):
+        return True
     if not _ensure_initialized(logger):
         return False
     normalized = _normalize_for_profanity(text)

tests/test_profanity_filter.py

@@ -36,6 +36,22 @@ class TestProfanityFilterEdgeCases:
     def test_contains_profanity_non_string_returns_false(self):
         assert contains_profanity(123) is False
 
+    def test_hate_symbol_swastika_detected(self):
+        """CJK swastika Unicode in text is detected as profanity (no better_profanity needed)."""
+        assert contains_profanity("\u5350") is True   # 卐
+        assert contains_profanity("\u534d") is True   # 卍
+        assert contains_profanity("User\u5350name") is True
+        assert contains_profanity("Hello \u534d world") is True
+
+    def test_hate_symbol_swastika_censored(self):
+        """CJK swastika Unicode is replaced with *** (no better_profanity needed)."""
+        assert censor("\u5350") == "***"
+        assert censor("\u534d") == "***"
+        assert censor("User\u5350name") == "User***name"
+        assert "***" in censor("Hello \u534d world")
+        assert "\u5350" not in censor("User\u5350name")
+        assert "\u534d" not in censor("Hello \u534d world")
+
 
 class TestProfanityFilterWithLibrary:
     """Tests that require better_profanity to be installed (skip if not)."""
@@ -125,3 +141,10 @@ class TestProfanityFilterFallbackWhenLibraryUnavailable:
                 censor("hello", logger=logger)
                 logger.warning.assert_called_once()
                 assert "better-profanity" in logger.warning.call_args[0][0]
+
+    def test_hate_symbol_still_detected_and_censored_when_library_unavailable(self):
+        """Hate symbols (e.g. swastika) are detected and replaced even when better_profanity is not installed."""
+        import modules.profanity_filter as pf
+        with patch.object(pf, "_profanity_available", False):
+            assert contains_profanity("\u5350") is True
+            assert censor("\u5350") == "***"