prevent XXS in third party apps

CommaFeed React client already strips malicious "javascript:" urls but we might as well not store those in the database

commafeed-server/src/main/java/com/commafeed/backend/Urls.java

@@ -8,6 +8,7 @@ import org.netpreserve.urlcanon.Canonicalizer;
 import org.netpreserve.urlcanon.ParsedUrl;
 
 import java.net.URI;
+import java.util.Locale;
 import java.util.regex.Pattern;
 
 @UtilityClass
@@ -17,17 +18,38 @@ public class Urls {
     private static final Pattern QUESTION_MARK = Pattern.compile(Pattern.quote("?"));
 
     public static boolean isHttp(String url) {
-        return url.startsWith("http://");
+        if (url == null) {
+            return false;
+        }
+
+        return url.toLowerCase(Locale.ROOT).startsWith("http://");
     }
 
     public static boolean isHttps(String url) {
-        return url.startsWith("https://");
+        if (url == null) {
+            return false;
+        }
+
+        return url.toLowerCase(Locale.ROOT).startsWith("https://");
     }
 
     public static boolean isAbsolute(String url) {
         return isHttp(url) || isHttps(url);
     }
 
+    /** remove malicious 'javascript: 'URLs * */
+    public static String sanitize(String url) {
+        if (url == null) {
+            return null;
+        }
+
+        if (!isHttp(url) && !isHttps(url)) {
+            return null;
+        }
+
+        return url;
+    }
+
     /**
      * @param relativeUrl the url of the entry
      * @param feedLink the url of the feed as described in the feed

commafeed-server/src/main/java/com/commafeed/backend/feed/parser/FeedParser.java

@@ -80,8 +80,8 @@ public class FeedParser {
             handleForeignMarkup(feed);
 
             String title = feed.getTitle();
-            String link = feed.getLink();
-            String iconUrl = feed.getIcon() != null ? feed.getIcon().getUrl() : null;
+            String link = Urls.sanitize(feed.getLink());
+            String iconUrl = Urls.sanitize(feed.getIcon() != null ? feed.getIcon().getUrl() : null);
             List<Entry> entries = buildEntries(feed, feedUrl);
             Instant lastEntryDate = entries.stream().findFirst().map(Entry::published).orElse(null);
             Instant lastPublishedDate = toValidInstant(feed.getPublishedDate(), false);
@@ -145,7 +145,7 @@ public class FeedParser {
             Instant publishedDate = buildEntryPublishedDate(item);
             Content content = buildContent(item);
 
-            entries.add(new Entry(guid, url, publishedDate, content));
+            entries.add(new Entry(guid, Urls.sanitize(url), publishedDate, content));
         }
 
         entries.sort(ENTRY_COMPARATOR);
@@ -173,7 +173,7 @@ public class FeedParser {
             return null;
         }
 
-        return new Enclosure(enclosure.getUrl(), enclosure.getType());
+        return new Enclosure(Urls.sanitize(enclosure.getUrl()), enclosure.getType());
     }
 
     private Instant buildEntryPublishedDate(SyndEntry item) {
@@ -277,7 +277,7 @@ public class FeedParser {
             return null;
         }
 
-        return new Media(description, thumbnailUrl, thumbnailWidth, thumbnailHeight);
+        return new Media(description, Urls.sanitize(thumbnailUrl), thumbnailWidth, thumbnailHeight);
     }
 
     private Long averageTimeBetweenEntries(List<Entry> entries) {

commafeed-server/src/test/java/com/commafeed/backend/UrlsTest.java

@@ -93,6 +93,35 @@ class UrlsTest {
                         "https://www.berliner-zeitung.de/feed.xml"));
     }
 
+    @Test
+    void testSanitize() {
+        // valid http/https urls are kept as is
+        Assertions.assertEquals("http://example.com/foo", Urls.sanitize("http://example.com/foo"));
+        Assertions.assertEquals(
+                "https://example.com/foo", Urls.sanitize("https://example.com/foo"));
+
+        // scheme matching is case-insensitive
+        Assertions.assertEquals("HTTP://example.com/foo", Urls.sanitize("HTTP://example.com/foo"));
+        Assertions.assertEquals(
+                "HTTPS://example.com/foo", Urls.sanitize("HTTPS://example.com/foo"));
+
+        // malicious or disallowed schemes are stripped
+        Assertions.assertNull(Urls.sanitize("javascript:alert(1)"));
+        Assertions.assertNull(Urls.sanitize("JavaScript:alert(1)"));
+        Assertions.assertNull(Urls.sanitize("data:text/html,<script>alert(1)</script>"));
+        Assertions.assertNull(Urls.sanitize("mailto:foo@example.com"));
+        Assertions.assertNull(Urls.sanitize("ftp://example.com/foo"));
+
+        // relative and protocol-relative urls are not considered valid
+        Assertions.assertNull(Urls.sanitize("/blog/entry/1"));
+        Assertions.assertNull(Urls.sanitize("//example.com/foo"));
+
+        // null and blank input
+        Assertions.assertNull(Urls.sanitize(null));
+        Assertions.assertNull(Urls.sanitize(""));
+        Assertions.assertNull(Urls.sanitize("   "));
+    }
+
     @Test
     void testRemoveTrailingSlash() {
         final String url = "http://localhost/";