diff --git a/sklearn/datasets/kddcup99.py b/sklearn/datasets/kddcup99.py
index 589749851ea6a..2b9686b4db2e0 100644
--- a/sklearn/datasets/kddcup99.py
+++ b/sklearn/datasets/kddcup99.py
@@ -24,7 +24,7 @@
 
 from .base import get_data_home
 from .base import Bunch
-from ..externals import joblib
+from ..externals import joblib, six
 from ..utils import check_random_state
 from ..utils import shuffle as shuffle_method
 
@@ -40,8 +40,8 @@
 
 
 def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
-                   percent10=False):
-    """Load and return the kddcup 99 dataset (regression).
+                   percent10=True, download_if_missing=True):
+    """Load and return the kddcup 99 dataset (classification).
 
     The KDD Cup '99 dataset was created by processing the tcpdump portions
     of the 1998 DARPA Intrusion Detection System (IDS) Evaluation dataset,
@@ -93,7 +93,7 @@ def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
 
     ================      ==========================================
     Samples total         699691
-    Dimensionality        40
+    Dimensionality        4
     Features              discrete (int) or continuous (float)
     Targets               str, 'normal.' or name of the anomaly type
     ================      ==========================================
@@ -102,7 +102,7 @@ def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
 
     ================      ==========================================
     Samples total         619052
-    Dimensionality        39
+    Dimensionality        3
     Features              discrete (int) or continuous (float)
     Targets               str, 'normal.' or name of the anomaly type
     ================      ==========================================
@@ -111,7 +111,7 @@ def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
 
     ================      ==========================================
     Samples total         95373
-    Dimensionality        39
+    Dimensionality        3
     Features              discrete (int) or continuous (float)
     Targets               str, 'normal.' or name of the anomaly type
     ================      ==========================================
@@ -135,6 +135,10 @@ def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
     percent10 : bool, default=False
         Whether to load only 10 percent of the data.
 
+    download_if_missing : bool, default=True
+        If False, raise a IOError if the data is not locally available
+        instead of trying to download the data from the source site.
+
     Returns
     -------
     data : Bunch
@@ -153,13 +157,14 @@ def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
            Intrusions in Unlabeled Data (2002) by Eleazar Eskin, Andrew Arnold,
            Michael Prerau, Leonid Portnoy, Sal Stolfo
     """
-    kddcup99 = _fetch_brute_kddcup99(shuffle=shuffle, percent10=percent10)
+    kddcup99 = _fetch_brute_kddcup99(shuffle=shuffle, percent10=percent10,
+                                     download_if_missing=download_if_missing)
 
     data = kddcup99.data
     target = kddcup99.target
 
     if subset == 'SA':
-        s = target == 'normal.'
+        s = target == b'normal.'
         t = np.logical_not(s)
         normal_samples = data[s, :]
         normal_targets = target[s]
@@ -187,13 +192,13 @@ def fetch_kddcup99(subset=None, shuffle=False, random_state=None,
         data[:, 5] = np.log((data[:, 5] + 0.1).astype(float))
 
         if subset == 'http':
-            s = data[:, 2] == 'http'
+            s = data[:, 2] == b'http'
             data = data[s]
             target = target[s]
             data = np.c_[data[:, 0], data[:, 4], data[:, 5]]
 
         if subset == 'smtp':
-            s = data[:, 2] == 'smtp'
+            s = data[:, 2] == b'smtp'
             data = data[s]
             target = target[s]
             data = np.c_[data[:, 0], data[:, 4], data[:, 5]]
@@ -320,6 +325,8 @@ def _fetch_brute_kddcup99(subset=None, data_home=None,
         file_ = GzipFile(fileobj=f, mode='r')
         Xy = []
         for line in file_.readlines():
+            if six.PY3:
+                line = line.decode()
             Xy.append(line.replace('\n', '').split(','))
         file_.close()
         print('extraction done')
diff --git a/sklearn/datasets/tests/test_kddcup99.py b/sklearn/datasets/tests/test_kddcup99.py
new file mode 100644
index 0000000000000..414c89763c1e8
--- /dev/null
+++ b/sklearn/datasets/tests/test_kddcup99.py
@@ -0,0 +1,41 @@
+"""Test  kddcup99 loader. Only 'percent10' mode is tested, as the full data
+is too big to use in unit-testing.
+
+The test is skipped if the data wasn't previously fetched and saved to
+scikit-learn data folder.
+"""
+
+import errno
+from sklearn.datasets import fetch_kddcup99
+from sklearn.utils.testing import assert_equal, SkipTest
+
+
+def test_percent10():
+    try:
+        data = fetch_kddcup99(download_if_missing=False)
+    except IOError as e:
+        if e.errno == errno.ENOENT:
+            raise SkipTest("kddcup99 dataset can not be loaded.")
+
+    assert_equal(data.data.shape, (494021, 41))
+    assert_equal(data.target.shape, (494021,))
+
+    data_shuffled = fetch_kddcup99(shuffle=True, random_state=0)
+    assert_equal(data.data.shape, data_shuffled.data.shape)
+    assert_equal(data.target.shape, data_shuffled.target.shape)
+
+    data = fetch_kddcup99('SA')
+    assert_equal(data.data.shape, (100655, 41))
+    assert_equal(data.target.shape, (100655,))
+
+    data = fetch_kddcup99('SF')
+    assert_equal(data.data.shape, (73237, 4))
+    assert_equal(data.target.shape, (73237,))
+
+    data = fetch_kddcup99('http')
+    assert_equal(data.data.shape, (58725, 3))
+    assert_equal(data.target.shape, (58725,))
+
+    data = fetch_kddcup99('smtp')
+    assert_equal(data.data.shape, (9571, 3))
+    assert_equal(data.target.shape, (9571,))