diff --git a/src/anonymeter/evaluators/inference_evaluator.py b/src/anonymeter/evaluators/inference_evaluator.py
index 51682ea..af716e2 100644
--- a/src/anonymeter/evaluators/inference_evaluator.py
+++ b/src/anonymeter/evaluators/inference_evaluator.py
@@ -324,7 +324,7 @@ def risk_for_groups(self, confidence_level: float = 0.95) -> dict[str, Evaluatio
                 # Count the number of success control attacks
                 n_control = evaluate_inference_guesses(
                     guesses=self._guesses_control.loc[common_indices],
-                    secrets=data_control[self._secret],
+                    secrets=data_control[self._secret].loc[common_indices],
                     regression=self._regression,
                 ).sum()
             else:
diff --git a/src/anonymeter/neighbors/mixed_types_kneighbors.py b/src/anonymeter/neighbors/mixed_types_kneighbors.py
index 318628a..be7c402 100644
--- a/src/anonymeter/neighbors/mixed_types_kneighbors.py
+++ b/src/anonymeter/neighbors/mixed_types_kneighbors.py
@@ -277,4 +277,7 @@ def predict(self, x: pd.DataFrame) -> pd.Series:
         guesses_idx = self._nn.kneighbors(queries=x[self._columns])
         if isinstance(guesses_idx, tuple):
             raise RuntimeError("guesses_idx cannot be a tuple")
-        return self._target_series.iloc[guesses_idx.flatten()]
+
+        guesses = self._target_series.iloc[guesses_idx.flatten()]
+        guesses.index = x.index
+        return guesses
diff --git a/tests/test_mixed_types_kneigbors.py b/tests/test_mixed_types_kneigbors.py
index 21e0b26..a7315b7 100644
--- a/tests/test_mixed_types_kneigbors.py
+++ b/tests/test_mixed_types_kneigbors.py
@@ -5,7 +5,7 @@
 import pandas as pd
 import pytest
 
-from anonymeter.neighbors.mixed_types_kneighbors import MixedTypeKNeighbors, gower_distance
+from anonymeter.neighbors.mixed_types_kneighbors import KNNInferencePredictor, MixedTypeKNeighbors, gower_distance
 
 from tests.fixtures import get_adult
 
@@ -79,3 +79,14 @@ def test_gower_distance_numerical():
     r0, r1 = rng.random(size=10), rng.random(size=10)
     dist = gower_distance(r0=r0, r1=r1, cat_cols_index=10)
     np.testing.assert_almost_equal(dist, np.sum(np.abs(r0 - r1)))
+
+
+def test_knn_inference_predictor_prediction_index_alignment():
+    df = get_adult("ori", n_samples=10)
+    aux_cols = ["age", "education", "sex"]
+    predictor = KNNInferencePredictor(data=df, columns=aux_cols, target_col="income", n_jobs=1)
+    queries = df[aux_cols]
+
+    guesses = predictor.predict(queries)
+
+    pd.testing.assert_index_equal(guesses.index, queries.index)