diff --git a/src/cryptoadvance/specter/helpers.py b/src/cryptoadvance/specter/helpers.py
index b5362fdf2..0dc9037c7 100644
--- a/src/cryptoadvance/specter/helpers.py
+++ b/src/cryptoadvance/specter/helpers.py
@@ -17,6 +17,7 @@ from embit.transaction import Transaction
 from embit.liquid.pset import PSET
 from embit.liquid.transaction import LTransaction
 from embit.liquid.networks import NETWORKS
+from urllib.parse import unquote
 from .persistence import read_json_file, write_json_file
 from .util.bcur import bcur_decode
 import threading
@@ -347,4 +348,22 @@ def get_address_from_dict(data_dict):
 
 
 def is_relative_url(url):
-    return re.match(r"^\/[^\/\\]", url)
+    # Decode percent-encoded characters so encoded control/whitespace cannot
+    # bypass validation, e.g. /%09/baidu.com -> "/\t/baidu.com".
+    normalized = unquote(url)
+
+    if not normalized or not normalized.startswith("/"):
+        return False
+    # Reject protocol-relative URLs after decoding.
+    if normalized.startswith("//"):
+        return False
+    # Reject any decoded control characters or whitespace anywhere in the URL,
+    # including ASCII DEL (0x7f), since browsers may normalize them in unsafe
+    # ways.
+    if any(ord(c) < 32 or ord(c) == 127 or c.isspace() for c in normalized):
+        return False
+    # Reject backslashes to avoid alternate path separator confusion.
+    if "\\" in normalized:
+        return False
+
+    return True
diff --git a/tests/test_helpers.py b/tests/test_helpers.py
index f508726ea..a1013544c 100644
--- a/tests/test_helpers.py
+++ b/tests/test_helpers.py
@@ -116,3 +116,44 @@ def test_create_unique_id():
     assert not all(
         create_unique_id(name) == "ghost_wallet" for name in not_the_same_unique_names
     )
+
+
+def test_is_relative_url():
+    """Test is_relative_url() for open redirect prevention."""
+    from cryptoadvance.specter.helpers import is_relative_url
+
+    # Valid relative URLs - should pass
+    assert is_relative_url("/")
+    assert is_relative_url("/valid/path")
+    assert is_relative_url("/path?query=value")
+    assert is_relative_url("/path#anchor")
+    assert is_relative_url("/a/b/c/d/e")
+
+    # Invalid - protocol-relative URLs (should be rejected)
+    assert not is_relative_url("//evil.com")
+
+    # Invalid - percent-encoded bypass attempts
+    assert not is_relative_url("/%09/baidu.com")  # tab -> // bypass
+    assert not is_relative_url("/%09baidu.com")
+    assert not is_relative_url("/%5c%5cevil.com")  # backslash encoding
+    assert not is_relative_url("/%20/path")  # space
+
+    # Invalid - control characters
+    assert not is_relative_url("/\x00/path")
+    assert not is_relative_url("/\x1f/path")
+    assert not is_relative_url("/\x7f/path")  # DEL character
+
+    # Invalid - whitespace
+    assert not is_relative_url("/ path")
+    assert not is_relative_url("/path ")  # trailing
+    assert not is_relative_url(" /path")  # leading
+
+    # Invalid - backslashes
+    assert not is_relative_url("/\\path")
+    assert not is_relative_url("/path\\to\\file")
+
+    # Invalid - non-relative
+    assert not is_relative_url("")
+    assert not is_relative_url("https://evil.com")
+    # /https:// passes validation - it's a local server path, not external
+    assert is_relative_url("/https://evil.com")