From 1d04be5c74d23ef084911d14339b336a5c2874fa Mon Sep 17 00:00:00 2001
From: Jake Jarvis <jake@jarv.is>
Date: Sat, 19 Sep 2026 23:04:01 -0400
Subject: [PATCH] fix: tighten `isPrivacyName` to avoid false positives on
 ambiguous privacy words

- Split keyword list into `PRIVACY_STRONG_KEYWORDS` (matched as plain substrings) and two weaker sets (`PRIVACY_WEAK_WORDS` / `PRIVACY_CONTEXT_WORDS`) that only trigger when at least two distinct terms appear together, preventing names like "Private Equity Partners LLC" or "Protection One" from being misclassified
- Remove empty-placeholder values (`-`, `n/a`, `none`, etc.) from the privacy check; they are not privacy notices and should be handled separately by callers
- Add `privacy.test.ts` covering redaction notices, known privacy-service names, ordinary names with ambiguous words, and empty placeholders
- Update README comment to clarify that `privacyEnabled` is set by name heuristics (privacy-service and redaction phrases) or RFC 9537 redactions
---
 README.md               |  2 +-
 src/lib/privacy.test.ts | 31 +++++++++++++++++++
 src/lib/privacy.ts      | 66 +++++++++++++++++++++++++++--------------
 3 files changed, 75 insertions(+), 24 deletions(-)
 create mode 100644 src/lib/privacy.test.ts

diff --git a/README.md b/README.md
index 1919a29..0103544 100644
--- a/README.md
+++ b/README.md
@@ -555,7 +555,7 @@ interface DomainRecord {
     country?: string;
     countryCode?: string;
   }>;
-  privacyEnabled?: boolean; // registrant appears privacy-redacted based on keyword heuristics or RFC 9537 redactions
+  privacyEnabled?: boolean; // registrant appears privacy-redacted based on name heuristics (privacy-service and redaction phrases) or RFC 9537 redactions
   redactions?: Array<{
     name: string; // e.g. "Registrant Email"
     prePath?: string;
diff --git a/src/lib/privacy.test.ts b/src/lib/privacy.test.ts
new file mode 100644
index 0000000..5cbb857
--- /dev/null
+++ b/src/lib/privacy.test.ts
@@ -0,0 +1,31 @@
+import { expect, test } from "vitest";
+import { isPrivacyName } from "./privacy";
+
+test("isPrivacyName flags redaction notices and privacy services", () => {
+  for (const v of [
+    "REDACTED FOR PRIVACY",
+    "Data Redacted",
+    "Privacy Protect, LLC",
+    "WhoisGuard Protected",
+    "Domains By Proxy, LLC",
+    "Private Registration",
+    "Domain Protection Services",
+    "Identity Protection Service",
+    "Withheld for Privacy ehf",
+    "Datos Privados",
+  ]) {
+    expect(isPrivacyName(v), v).toBe(true);
+  }
+});
+
+test("isPrivacyName does not flag ordinary names with ambiguous words", () => {
+  for (const v of ["Private Equity Partners LLC", "Protection One", "Jane Private", "Acme Inc"]) {
+    expect(isPrivacyName(v), v).toBe(false);
+  }
+});
+
+test("isPrivacyName does not treat empty placeholders as privacy", () => {
+  for (const v of ["-", "n/a", "N/A", "none", "not available"]) {
+    expect(isPrivacyName(v), v).toBe(false);
+  }
+});
diff --git a/src/lib/privacy.ts b/src/lib/privacy.ts
index 1cd3879..27d608d 100644
--- a/src/lib/privacy.ts
+++ b/src/lib/privacy.ts
@@ -1,41 +1,61 @@
-export const PRIVACY_NAME_KEYWORDS = [
-  "redacted",
-  "privacy",
-  "private",
+/**
+ * Phrases that indicate privacy/redaction on their own. Matched as case-insensitive substrings.
+ */
+export const PRIVACY_STRONG_KEYWORDS = [
+  "redacted", // also covers "redacted for privacy", "redacted.forprivacy"
   "withheld",
   "not disclosed",
-  "protected",
-  "protection",
   "privado", // Spanish
   "datos privados", // Spanish
   "data protected",
-  "data redacted",
-  "gdpr redacted",
   "gdpr masked",
   "non-public data",
   "statutory masking",
-  "redacted.forprivacy",
   "registration private",
+  "private registration",
   "hidden upon user request",
   "not available from registry",
+  "whois privacy",
+  "whoisguard",
+  "privacy protect",
+  "privacy service",
+  "domain privacy",
+  "contact privacy",
+  "domains by proxy",
+  "proxy service",
+  "for privacy",
 ];
 
-// Completely unusable/empty values that should be filtered
-export const NO_DATA_VALUES = [
-  "-",
-  ".",
-  "n/a",
-  "na",
-  "no data",
-  "not available",
-  "not applicable",
-  "none",
+/**
+ * Words too ambiguous to trust alone ("Private Equity LLC", "Protection One"). They only count
+ * when at least two distinct terms from WEAK + CONTEXT appear as whole words.
+ */
+const PRIVACY_WEAK_WORDS = ["privacy", "private", "protect", "protected", "protection"];
+const PRIVACY_CONTEXT_WORDS = [
+  "whois",
+  "proxy",
+  "domain",
+  "domains",
+  "registration",
+  "guard",
+  "service",
+  "services",
+  "masked",
+  "anonymous",
+  "identity",
+  "contact",
 ];
+const WEAK_WORD_RE = new RegExp(`\\b(?:${PRIVACY_WEAK_WORDS.join("|")})\\b`, "g");
+const CONTEXT_WORD_RE = new RegExp(
+  `\\b(?:${[...PRIVACY_WEAK_WORDS, ...PRIVACY_CONTEXT_WORDS].join("|")})\\b`,
+  "g",
+);
 
+/** True when a registrant name/organization looks like a privacy service or redaction notice. */
 export function isPrivacyName(value: string): boolean {
   const v = value.toLowerCase().trim();
-  // Check for complete no-data values
-  if (NO_DATA_VALUES.includes(v)) return true;
-  // Check for privacy keywords
-  return PRIVACY_NAME_KEYWORDS.some((k) => v.includes(k));
+  if (PRIVACY_STRONG_KEYWORDS.some((k) => v.includes(k))) return true;
+  if (!WEAK_WORD_RE.test(v)) return false;
+  WEAK_WORD_RE.lastIndex = 0;
+  return new Set(v.match(CONTEXT_WORD_RE)).size >= 2;
 }