diff --git a/bot_bottle/dlp_detectors.py b/bot_bottle/dlp_detectors.py index a9603db..35d9312 100644 --- a/bot_bottle/dlp_detectors.py +++ b/bot_bottle/dlp_detectors.py @@ -112,19 +112,33 @@ JAILBREAK_PHRASES: tuple[re.Pattern[str], ...] = ( PROXIMITY_CHARS = 500 -def _min_distance( +def _nearby( a_matches: list[re.Match[str]], b_matches: list[re.Match[str]], + threshold: int, ) -> int | None: - """Smallest char distance between any pair of matches.""" + """Return the smallest char gap between any a–b pair, or None if + both lists are empty. O(n log n) via sort + linear scan.""" if not a_matches or not b_matches: return None - best = None - for a in a_matches: - for b in b_matches: - gap = max(0, max(a.start(), b.start()) - min(a.end(), b.end())) + events = sorted( + [(m.start(), m.end(), "a") for m in a_matches] + + [(m.start(), m.end(), "b") for m in b_matches], + ) + best: int | None = None + prev_end: int | None = None + prev_tag: str | None = None + for start, end, tag in events: + if prev_tag is not None and prev_tag != tag and prev_end is not None: + gap = max(0, start - prev_end) if best is None or gap < best: best = gap + if best == 0: + return 0 + if best <= threshold: + return best + prev_end = end if prev_end is None else max(prev_end, end) + prev_tag = tag return best @@ -133,7 +147,7 @@ def scan_naive_injection(text: str) -> ScanResult | None: jailbreak_hits = [m for p in JAILBREAK_PHRASES for m in p.finditer(text)] if disclosure_hits and jailbreak_hits: - dist = _min_distance(disclosure_hits, jailbreak_hits) + dist = _nearby(disclosure_hits, jailbreak_hits, PROXIMITY_CHARS) if dist is not None and dist <= PROXIMITY_CHARS: return ScanResult( severity="block",