diff --git a/android/guava-tests/test/com/google/common/io/BaseEncodingTest.java b/android/guava-tests/test/com/google/common/io/BaseEncodingTest.java
index a07cfccadb7c..96f419945b6b 100644
--- a/android/guava-tests/test/com/google/common/io/BaseEncodingTest.java
+++ b/android/guava-tests/test/com/google/common/io/BaseEncodingTest.java
@@ -150,6 +150,18 @@ public void testBase64InvalidDecodings() {
     assertFailsToDecode(base64(), "?", "Invalid input length 1");
   }
 
+  public void testBase64RejectsNonCanonicalPaddingBits() {
+    // Bits of the final character that are not part of any decoded byte must be zero (RFC 4648
+    // section 3.5); otherwise many distinct strings decode to the same bytes.
+    assertFailsToDecode(base64(), "AB==", "Unused bits set in encoded input");
+    assertFailsToDecode(base64(), "QE==", "Unused bits set in encoded input");
+    assertFailsToDecode(base64(), "AAB=", "Unused bits set in encoded input");
+    // The canonical siblings still decode.
+    testDecodesByBytes(base64(), "AA==", new byte[] {0x00});
+    testDecodesByBytes(base64(), "QA==", new byte[] {0x40});
+    testDecodesByBytes(base64(), "AAA=", new byte[] {0x00, 0x00});
+  }
+
   public void testBase64CannotUpperCase() {
     assertThrows(IllegalStateException.class, () -> base64().upperCase());
   }
@@ -226,6 +238,10 @@ public void testBase64UrlInvalidDecodings() {
     assertFailsToDecode(base64Url(), "/zzz", "Unrecognized character: /");
   }
 
+  public void testBase64UrlRejectsNonCanonicalPaddingBits() {
+    assertFailsToDecode(base64Url(), "AB", "Unused bits set in encoded input");
+  }
+
   public void testBase32() {
     // The following test vectors are specified in RFC 4648 itself
     testEncodingWithCasing(base32(), "", "");
@@ -285,6 +301,15 @@ public void testBase32InvalidDecodings() {
     assertFailsToDecode(base32(), "?", "Invalid input length 1");
   }
 
+  public void testBase32RejectsNonCanonicalPaddingBits() {
+    // Bits of the final character that are not part of any decoded byte must be zero (RFC 4648
+    // section 3.5); "MZXW7" and "MZXW6" would otherwise both decode to "foo".
+    assertFailsToDecode(base32(), "MZXW7===", "Unused bits set in encoded input");
+    assertFailsToDecode(base32(), "AAAB====", "Unused bits set in encoded input");
+    testDecodes(base32(), "MZXW6===", "foo");
+    testDecodesByBytes(base32(), "AAAA====", new byte[] {0x00, 0x00});
+  }
+
   public void testBase32UpperCaseIsNoOp() {
     assertThat(base32().upperCase()).isSameInstanceAs(base32());
   }
@@ -353,6 +378,12 @@ public void testBase32HexInvalidDecodings() {
     assertFailsToDecode(base32Hex(), "ABCDEF");
   }
 
+  public void testBase32HexRejectsNonCanonicalPaddingBits() {
+    assertFailsToDecode(base32Hex(), "CPNMV===", "Unused bits set in encoded input");
+    assertFailsToDecode(base32Hex(), "0001====", "Unused bits set in encoded input");
+    testDecodes(base32Hex(), "CPNMU===", "foo");
+  }
+
   public void testBase32HexUpperCaseIsNoOp() {
     assertThat(base32Hex().upperCase()).isSameInstanceAs(base32Hex());
   }
diff --git a/android/guava/src/com/google/common/io/BaseEncoding.java b/android/guava/src/com/google/common/io/BaseEncoding.java
index 0f311b38ecb8..4e191bad8f1a 100644
--- a/android/guava/src/com/google/common/io/BaseEncoding.java
+++ b/android/guava/src/com/google/common/io/BaseEncoding.java
@@ -753,6 +753,20 @@ public boolean canDecode(CharSequence chars) {
           return false;
         }
       }
+      // Any bits of the final character not consumed by a decoded byte must be zero for a
+      // canonical encoding (RFC 4648 section 3.5).
+      int lastChunkChars = chars.length() % alphabet.charsPerChunk;
+      if (lastChunkChars != 0) {
+        int unusedBits = (lastChunkChars * alphabet.bitsPerChar) % 8;
+        try {
+          if ((alphabet.decode(chars.charAt(chars.length() - 1)) & ((1 << unusedBits) - 1)) != 0) {
+            return false;
+          }
+        } catch (DecodingException e) {
+          // Can't happen: the loop above already verified every character is decodable.
+          return false;
+        }
+      }
       return true;
     }
 
@@ -774,9 +788,15 @@ int decodeTo(byte[] target, CharSequence chars) throws DecodingException {
           }
         }
         int minOffset = alphabet.bytesPerChunk * 8 - charsProcessed * alphabet.bitsPerChar;
-        for (int offset = (alphabet.bytesPerChunk - 1) * 8; offset >= minOffset; offset -= 8) {
+        int offset = (alphabet.bytesPerChunk - 1) * 8;
+        for (; offset >= minOffset; offset -= 8) {
           target[bytesWritten++] = (byte) ((chunk >>> offset) & 0xFF);
         }
+        // Any bits of the final character not consumed by an emitted byte must be zero for a
+        // canonical encoding (RFC 4648 section 3.5).
+        if ((chunk & ((1L << (offset + 8)) - 1)) != 0) {
+          throw new DecodingException("Unused bits set in encoded input");
+        }
       }
       return bytesWritten;
     }
@@ -800,6 +820,11 @@ public int read() throws IOException {
               if (!hitPadding && !alphabet.isValidPaddingStartPosition(readChars)) {
                 throw new DecodingException("Invalid input length " + readChars);
               }
+              // Any trailing bits not consumed by an emitted byte must be zero for a canonical
+              // encoding (RFC 4648 section 3.5).
+              if ((bitBuffer & ((1 << bitBufferLength) - 1)) != 0) {
+                throw new DecodingException("Unused bits set in encoded input");
+              }
               return -1;
             }
             readChars++;
@@ -1045,7 +1070,13 @@ int decodeTo(byte[] target, CharSequence chars) throws DecodingException {
           if (i < chars.length()) {
             chunk |= alphabet.decode(chars.charAt(i++));
             target[bytesWritten++] = (byte) (chunk & 0xFF);
+          } else if ((chunk & 0xFF) != 0) {
+            // Unused bits of the third character must be zero (RFC 4648 section 3.5).
+            throw new DecodingException("Unused bits set in encoded input");
           }
+        } else if ((chunk & 0xFFFF) != 0) {
+          // Unused bits of the second character must be zero (RFC 4648 section 3.5).
+          throw new DecodingException("Unused bits set in encoded input");
         }
       }
       return bytesWritten;
diff --git a/guava-tests/test/com/google/common/io/BaseEncodingTest.java b/guava-tests/test/com/google/common/io/BaseEncodingTest.java
index a07cfccadb7c..96f419945b6b 100644
--- a/guava-tests/test/com/google/common/io/BaseEncodingTest.java
+++ b/guava-tests/test/com/google/common/io/BaseEncodingTest.java
@@ -150,6 +150,18 @@ public void testBase64InvalidDecodings() {
     assertFailsToDecode(base64(), "?", "Invalid input length 1");
   }
 
+  public void testBase64RejectsNonCanonicalPaddingBits() {
+    // Bits of the final character that are not part of any decoded byte must be zero (RFC 4648
+    // section 3.5); otherwise many distinct strings decode to the same bytes.
+    assertFailsToDecode(base64(), "AB==", "Unused bits set in encoded input");
+    assertFailsToDecode(base64(), "QE==", "Unused bits set in encoded input");
+    assertFailsToDecode(base64(), "AAB=", "Unused bits set in encoded input");
+    // The canonical siblings still decode.
+    testDecodesByBytes(base64(), "AA==", new byte[] {0x00});
+    testDecodesByBytes(base64(), "QA==", new byte[] {0x40});
+    testDecodesByBytes(base64(), "AAA=", new byte[] {0x00, 0x00});
+  }
+
   public void testBase64CannotUpperCase() {
     assertThrows(IllegalStateException.class, () -> base64().upperCase());
   }
@@ -226,6 +238,10 @@ public void testBase64UrlInvalidDecodings() {
     assertFailsToDecode(base64Url(), "/zzz", "Unrecognized character: /");
   }
 
+  public void testBase64UrlRejectsNonCanonicalPaddingBits() {
+    assertFailsToDecode(base64Url(), "AB", "Unused bits set in encoded input");
+  }
+
   public void testBase32() {
     // The following test vectors are specified in RFC 4648 itself
     testEncodingWithCasing(base32(), "", "");
@@ -285,6 +301,15 @@ public void testBase32InvalidDecodings() {
     assertFailsToDecode(base32(), "?", "Invalid input length 1");
   }
 
+  public void testBase32RejectsNonCanonicalPaddingBits() {
+    // Bits of the final character that are not part of any decoded byte must be zero (RFC 4648
+    // section 3.5); "MZXW7" and "MZXW6" would otherwise both decode to "foo".
+    assertFailsToDecode(base32(), "MZXW7===", "Unused bits set in encoded input");
+    assertFailsToDecode(base32(), "AAAB====", "Unused bits set in encoded input");
+    testDecodes(base32(), "MZXW6===", "foo");
+    testDecodesByBytes(base32(), "AAAA====", new byte[] {0x00, 0x00});
+  }
+
   public void testBase32UpperCaseIsNoOp() {
     assertThat(base32().upperCase()).isSameInstanceAs(base32());
   }
@@ -353,6 +378,12 @@ public void testBase32HexInvalidDecodings() {
     assertFailsToDecode(base32Hex(), "ABCDEF");
   }
 
+  public void testBase32HexRejectsNonCanonicalPaddingBits() {
+    assertFailsToDecode(base32Hex(), "CPNMV===", "Unused bits set in encoded input");
+    assertFailsToDecode(base32Hex(), "0001====", "Unused bits set in encoded input");
+    testDecodes(base32Hex(), "CPNMU===", "foo");
+  }
+
   public void testBase32HexUpperCaseIsNoOp() {
     assertThat(base32Hex().upperCase()).isSameInstanceAs(base32Hex());
   }
diff --git a/guava/src/com/google/common/io/BaseEncoding.java b/guava/src/com/google/common/io/BaseEncoding.java
index 0f311b38ecb8..4e191bad8f1a 100644
--- a/guava/src/com/google/common/io/BaseEncoding.java
+++ b/guava/src/com/google/common/io/BaseEncoding.java
@@ -753,6 +753,20 @@ public boolean canDecode(CharSequence chars) {
           return false;
         }
       }
+      // Any bits of the final character not consumed by a decoded byte must be zero for a
+      // canonical encoding (RFC 4648 section 3.5).
+      int lastChunkChars = chars.length() % alphabet.charsPerChunk;
+      if (lastChunkChars != 0) {
+        int unusedBits = (lastChunkChars * alphabet.bitsPerChar) % 8;
+        try {
+          if ((alphabet.decode(chars.charAt(chars.length() - 1)) & ((1 << unusedBits) - 1)) != 0) {
+            return false;
+          }
+        } catch (DecodingException e) {
+          // Can't happen: the loop above already verified every character is decodable.
+          return false;
+        }
+      }
       return true;
     }
 
@@ -774,9 +788,15 @@ int decodeTo(byte[] target, CharSequence chars) throws DecodingException {
           }
         }
         int minOffset = alphabet.bytesPerChunk * 8 - charsProcessed * alphabet.bitsPerChar;
-        for (int offset = (alphabet.bytesPerChunk - 1) * 8; offset >= minOffset; offset -= 8) {
+        int offset = (alphabet.bytesPerChunk - 1) * 8;
+        for (; offset >= minOffset; offset -= 8) {
           target[bytesWritten++] = (byte) ((chunk >>> offset) & 0xFF);
         }
+        // Any bits of the final character not consumed by an emitted byte must be zero for a
+        // canonical encoding (RFC 4648 section 3.5).
+        if ((chunk & ((1L << (offset + 8)) - 1)) != 0) {
+          throw new DecodingException("Unused bits set in encoded input");
+        }
       }
       return bytesWritten;
     }
@@ -800,6 +820,11 @@ public int read() throws IOException {
               if (!hitPadding && !alphabet.isValidPaddingStartPosition(readChars)) {
                 throw new DecodingException("Invalid input length " + readChars);
               }
+              // Any trailing bits not consumed by an emitted byte must be zero for a canonical
+              // encoding (RFC 4648 section 3.5).
+              if ((bitBuffer & ((1 << bitBufferLength) - 1)) != 0) {
+                throw new DecodingException("Unused bits set in encoded input");
+              }
               return -1;
             }
             readChars++;
@@ -1045,7 +1070,13 @@ int decodeTo(byte[] target, CharSequence chars) throws DecodingException {
           if (i < chars.length()) {
             chunk |= alphabet.decode(chars.charAt(i++));
             target[bytesWritten++] = (byte) (chunk & 0xFF);
+          } else if ((chunk & 0xFF) != 0) {
+            // Unused bits of the third character must be zero (RFC 4648 section 3.5).
+            throw new DecodingException("Unused bits set in encoded input");
           }
+        } else if ((chunk & 0xFFFF) != 0) {
+          // Unused bits of the second character must be zero (RFC 4648 section 3.5).
+          throw new DecodingException("Unused bits set in encoded input");
         }
       }
       return bytesWritten;