fix(security): consolidation & securite phase 1 (#84, BUG-021 a BUG-034)

- sanitizer XSS serveur (markdown + page de partage) [BUG-021/022]
- rate-limit/lockout MFA [BUG-023]
- isolation vaults par segments [BUG-024]
- caps regex ReDoS [BUG-025]
- SSRF webhooks + secrets externalises [BUG-026]
- rotation/revocation des jetons [BUG-027]
- politique de mot de passe + invalidation sessions [BUG-028]
- verrous users.json [BUG-029]
- IP reelle dans les audits [BUG-030]
- rate-limit par compte [BUG-031]
- symlinks hors vault ignores [BUG-032]
- recherche simple via inverted index [BUG-033]
- token en memoire + cookie HttpOnly, CSP durcie [BUG-034]

Tests: pytest 961 passed / 6 skipped, ruff 0, mypy 0, frontend vert.

backend/search.py

@@ -13,6 +13,11 @@ from sortedcontainers import SortedList
 from backend import indexer as _indexer
 from backend import semantic_search as _semantic
 from backend.indexer import index
+from backend.services.regex_safety import (
+    MAX_REGEX_MATCHES,
+    truncate_for_regex,
+    validate_regex,
+)
 
 logger = logging.getLogger("obsigate.search")
 
@@ -226,12 +231,16 @@ def _extract_regex_snippet(
     if not content or not pattern_text:
         return content[:200].strip() if content else ""
 
+    # BUG-025: bound the text scanned and the number of matches collected.
+    content = truncate_for_regex(content)
+
     try:
+        validate_regex(pattern_text)
         pattern = re.compile(pattern_text, re.IGNORECASE)
-    except re.error:
+    except (re.error, ValueError):
         return _escape_html(content[:200].strip())
 
-    matches = list(pattern.finditer(content))
+    matches = list(pattern.finditer(content))[:MAX_REGEX_MATCHES]
     if not matches:
         return _escape_html(content[:200].strip())
 
@@ -729,60 +738,97 @@ def search(
     query_lower = query.lower()
     results: list[dict[str, Any]] = []
 
-    for vault_name, vault_data in index.items():
-        if vault_filter != "all" and vault_name != vault_filter:
+    inv = get_inverted_index()
+    use_index = (not inv.is_stale()) and inv.doc_count > 0
+
+    if use_index:
+        # BUG-033: retrieve candidates from the inverted index instead of
+        # scanning every document.  Multi-term queries require all terms
+        # (a superset of exact-phrase matches), single terms use prefix
+        # expansion.  Falls back to a full scan while the index is building.
+        if has_query:
+            terms = [t for t in tokenize(query) if t]
+            if not terms:
+                return []
+            doc_sets: list[set] = []
+            for term in terms:
+                term_docs: set = set(inv.word_index.get(term, {}).keys())
+                if len(term) >= MIN_PREFIX_LENGTH:
+                    for expanded in inv.get_prefix_tokens(term):
+                        term_docs.update(inv.word_index.get(expanded, {}).keys())
+                doc_sets.append(term_docs)
+            doc_keys = set.intersection(*doc_sets) if doc_sets else set()
+        else:
+            doc_keys = set(inv.doc_info.keys())
+
+        if vault_filter != "all":
+            doc_keys &= inv.vault_docs.get(vault_filter, set())
+        for tag in selected_tags:
+            doc_keys &= inv.tag_docs.get(tag.lower(), set())
+
+        candidates = [
+            (inv.doc_vault[dk], inv.doc_info[dk])
+            for dk in doc_keys
+            if dk in inv.doc_info
+        ]
+    else:
+        candidates = [
+            (vault_name, file_info)
+            for vault_name, vault_data in index.items()
+            if vault_filter == "all" or vault_name == vault_filter
+            for file_info in vault_data["files"]
+        ]
+
+    for vault_name, file_info in candidates:
+        # Tag filter: all selected tags must be present
+        if selected_tags and not all(tag in file_info["tags"] for tag in selected_tags):
             continue
 
-        for file_info in vault_data["files"]:
-            # Tag filter: all selected tags must be present
-            if selected_tags and not all(tag in file_info["tags"] for tag in selected_tags):
-                continue
+        score = 0
+        snippet = file_info.get("content_preview", "")
 
-            score = 0
-            snippet = file_info.get("content_preview", "")
+        if has_query:
+            title_lower = file_info["title"].lower()
 
-            if has_query:
-                title_lower = file_info["title"].lower()
+            # Exact title match (highest weight)
+            if query_lower == title_lower:
+                score += 20
+            # Partial title match
+            elif query_lower in title_lower:
+                score += 10
 
-                # Exact title match (highest weight)
-                if query_lower == title_lower:
-                    score += 20
-                # Partial title match
-                elif query_lower in title_lower:
-                    score += 10
+            # Path match (folder/filename relevance)
+            if query_lower in file_info["path"].lower():
+                score += 5
 
-                # Path match (folder/filename relevance)
-                if query_lower in file_info["path"].lower():
-                    score += 5
+            # Tag name match
+            for tag in file_info.get("tags", []):
+                if query_lower in tag.lower():
+                    score += 3
+                    break  # count once per file
 
-                # Tag name match
-                for tag in file_info.get("tags", []):
-                    if query_lower in tag.lower():
-                        score += 3
-                        break  # count once per file
+            # Content match — use cached content (no disk I/O)
+            content = file_info.get("content", "")
+            content_lower = content.lower()
+            if query_lower in content_lower:
+                # Frequency-based scoring, capped to avoid over-weighting
+                occurrences = content_lower.count(query_lower)
+                score += min(occurrences, 10)
+                snippet = _extract_snippet(content, query)
+        else:
+            # Tag-only filter: all matching files get score 1
+            score = 1
 
-                # Content match — use cached content (no disk I/O)
-                content = file_info.get("content", "")
-                content_lower = content.lower()
-                if query_lower in content_lower:
-                    # Frequency-based scoring, capped to avoid over-weighting
-                    occurrences = content_lower.count(query_lower)
-                    score += min(occurrences, 10)
-                    snippet = _extract_snippet(content, query)
-            else:
-                # Tag-only filter: all matching files get score 1
-                score = 1
-
-            if score > 0:
-                results.append({
-                    "vault": vault_name,
-                    "path": file_info["path"],
-                    "title": file_info["title"],
-                    "tags": file_info["tags"],
-                    "score": score,
-                    "snippet": snippet,
-                    "modified": file_info["modified"],
-                })
+        if score > 0:
+            results.append({
+                "vault": vault_name,
+                "path": file_info["path"],
+                "title": file_info["title"],
+                "tags": file_info["tags"],
+                "score": score,
+                "snippet": snippet,
+                "modified": file_info["modified"],
+            })
 
     results.sort(key=lambda x: -x["score"])
     return results[:limit]
@@ -908,12 +954,14 @@ def _passes_search_filters(
     title = file_info.get("title", "")
     content = file_info.get("content", "")
     path = file_info.get("path", "")
-    search_text = f"{title} {content}"
+    # BUG-025: cap the text scanned by a user-supplied regex.
+    search_text = truncate_for_regex(f"{title} {content}")
     search_text_norm = normalize_text(search_text)
 
     # --- Regex mode ---
     if regex and raw_query:
         try:
+            validate_regex(raw_query)
             flags = 0 if case_sensitive else re.IGNORECASE
             if whole_word:
                 pattern = re.compile(rf"\b{raw_query}\b", flags)
@@ -921,7 +969,7 @@ def _passes_search_filters(
                 pattern = re.compile(raw_query, flags)
             if not pattern.search(search_text):
                 return False
-        except re.error:
+        except (re.error, ValueError):
             return False
         return _passes_path_filters(path, include_paths, exclude_paths)
 
@@ -1138,6 +1186,11 @@ def advanced_search(
     """
     t0 = time.monotonic()
     query = query.strip() if query else ""
+
+    # BUG-025: reject oversized / catastrophic regex patterns up front.
+    if regex and query:
+        validate_regex(query)
+
     parsed = _parse_advanced_query(query)
 
     # Merge explicit tag_filter with parsed tag: operators