diff --git a/decnet/clustering/impl/connected_components.py b/decnet/clustering/impl/connected_components.py
index 811f4343..3f511f1f 100644
--- a/decnet/clustering/impl/connected_components.py
+++ b/decnet/clustering/impl/connected_components.py
@@ -43,10 +43,15 @@ from decnet.clustering.impl.similarity import (
 )
 from decnet.logging import get_logger
 from decnet.profiler.identity_rollup import extract_fp_summaries
+from decnet.util.simhash import from_bytes8, to_bytes8
 from decnet.web.db.repository import BaseRepository
 
 log = get_logger("clustering.connected_components")
 
+# Per-session SimHash observations of the keystroke-rhythm biometric; the
+# rollup folds them into one identity-level centroid.
+_DIGRAPH_PRIMITIVE = "motor.digraph_simhash"
+
 
 def cluster_observations(
     observations: Iterable[Observation],
@@ -354,6 +359,38 @@ async def _link(
         return False
 
 
+async def _digraph_centroid(
+    repo: BaseRepository, identity_uuid: str,
+) -> Optional[bytes]:
+    """Fold the identity's session-level ``motor.digraph_simhash``
+    observations into one 8-byte bitwise-majority centroid.
+
+    Bit *i* is set iff a majority of the identity's session SimHashes
+    have it set — denoises per-session jitter so the centroid is the
+    stable keystroke-rhythm fingerprint for Hamming comparison. Returns
+    ``None`` when the identity has no usable digraph observations yet.
+    """
+    obs = await repo.observations_for_identity_primitive(
+        identity_uuid, _DIGRAPH_PRIMITIVE,
+    )
+    hashes: list[int] = []
+    for o in obs:
+        value = o.get("value")
+        if isinstance(value, str) and len(value) == 16:
+            try:
+                hashes.append(from_bytes8(bytes.fromhex(value)))
+            except ValueError:
+                continue
+    if not hashes:
+        return None
+    n = len(hashes)
+    centroid = 0
+    for i in range(64):
+        if sum((h >> i) & 1 for h in hashes) * 2 > n:
+            centroid |= (1 << i)
+    return to_bytes8(centroid)
+
+
 async def _roll_up_fingerprints(
     repo: BaseRepository,
     identity_uuid: str,
@@ -365,8 +402,11 @@ async def _roll_up_fingerprints(
     next pass."""
     summaries = extract_fp_summaries(member_rows)
     fp_kwargs = {k: v for k, v in summaries.items() if k in {"ja3_hashes", "hassh_hashes", "tls_cert_sha256"}}
+    kd_centroid = await _digraph_centroid(repo, identity_uuid)
     try:
-        await repo.update_identity_fingerprints(identity_uuid, **fp_kwargs)
+        await repo.update_identity_fingerprints(
+            identity_uuid, kd_digraph_simhash=kd_centroid, **fp_kwargs,
+        )
     except Exception:  # noqa: BLE001
         log.exception(
             "clusterer: failed to roll up fingerprints for identity=%s",
diff --git a/decnet/web/db/repository.py b/decnet/web/db/repository.py
index 7325b1f4..56f89e84 100644
--- a/decnet/web/db/repository.py
+++ b/decnet/web/db/repository.py
@@ -683,15 +683,19 @@ class BaseRepository(ABC):
         ja3_hashes: Optional[str] = None,
         hassh_hashes: Optional[str] = None,
         tls_cert_sha256: Optional[str] = None,
+        kd_digraph_simhash: Optional[bytes] = None,
     ) -> None:
         """Set the fingerprint summary columns on one ``AttackerIdentity``.
 
-        Each argument is a JSON-encoded ``list[str]`` (the federation
-        wire shape) or ``None`` to leave the corresponding column at
-        ``NULL``. Always overwrites — the rollup writer is the source
-        of truth for these columns, computed deterministically from
-        the identity's member observations every clusterer tick. Also
-        bumps ``updated_at`` so cache subscribers can invalidate.
+        ``ja3_hashes`` / ``hassh_hashes`` / ``tls_cert_sha256`` are
+        JSON-encoded ``list[str]`` (the federation wire shape) or
+        ``None``. ``kd_digraph_simhash`` is the 8-byte keystroke-rhythm
+        centroid (bitwise majority over the identity's session-level
+        ``motor.digraph_simhash`` observations) or ``None``. Always
+        overwrites — the rollup writer is the source of truth for these
+        columns, recomputed deterministically from the identity's member
+        observations every clusterer tick. Also bumps ``updated_at`` so
+        cache subscribers can invalidate.
         """
         pass
 
diff --git a/decnet/web/db/sqlmodel_repo/campaigns.py b/decnet/web/db/sqlmodel_repo/campaigns.py
index 78c04273..00f21bfe 100644
--- a/decnet/web/db/sqlmodel_repo/campaigns.py
+++ b/decnet/web/db/sqlmodel_repo/campaigns.py
@@ -105,6 +105,7 @@ class CampaignsMixin(_MixinBase):
             AttackerIdentity.hassh_hashes,
             AttackerIdentity.payload_simhashes,
             AttackerIdentity.c2_endpoints,
+            AttackerIdentity.kd_digraph_simhash,
         ).order_by(AttackerIdentity.created_at)
         if limit is not None:
             statement = statement.limit(limit)
@@ -129,6 +130,7 @@ class CampaignsMixin(_MixinBase):
                     "hassh_hashes": row.hassh_hashes,
                     "payload_simhashes": row.payload_simhashes,
                     "c2_endpoints": row.c2_endpoints,
+                    "kd_digraph_simhash": row.kd_digraph_simhash,
                 }
                 for row in result.all()
             ]
diff --git a/decnet/web/db/sqlmodel_repo/identities.py b/decnet/web/db/sqlmodel_repo/identities.py
index 6b9ca639..321c469a 100644
--- a/decnet/web/db/sqlmodel_repo/identities.py
+++ b/decnet/web/db/sqlmodel_repo/identities.py
@@ -173,6 +173,7 @@ class IdentitiesMixin(_MixinBase):
         ja3_hashes: Optional[str] = None,
         hassh_hashes: Optional[str] = None,
         tls_cert_sha256: Optional[str] = None,
+        kd_digraph_simhash: Optional[bytes] = None,
     ) -> None:
         statement = (
             update(AttackerIdentity)
@@ -181,6 +182,7 @@ class IdentitiesMixin(_MixinBase):
                 ja3_hashes=ja3_hashes,
                 hassh_hashes=hassh_hashes,
                 tls_cert_sha256=tls_cert_sha256,
+                kd_digraph_simhash=kd_digraph_simhash,
                 updated_at=datetime.now(timezone.utc),
             )
         )
diff --git a/tests/clustering/test_kd_digraph_centroid.py b/tests/clustering/test_kd_digraph_centroid.py
new file mode 100644
index 00000000..d85e531c
--- /dev/null
+++ b/tests/clustering/test_kd_digraph_centroid.py
@@ -0,0 +1,47 @@
+# SPDX-License-Identifier: AGPL-3.0-or-later
+"""``_digraph_centroid`` — bitwise-majority rollup of session SimHashes."""
+from __future__ import annotations
+
+from decnet.clustering.impl.connected_components import _digraph_centroid
+from decnet.util.simhash import from_bytes8, to_bytes8
+
+_ALL_ONES = (1 << 64) - 1
+
+
+class _FakeRepo:
+    """Returns canned digraph observations for one identity."""
+
+    def __init__(self, hash_ints: list[int]) -> None:
+        self._values = [to_bytes8(h).hex() for h in hash_ints]
+
+    async def observations_for_identity_primitive(self, identity_uuid, primitive):
+        assert primitive == "motor.digraph_simhash"
+        return [{"value": v} for v in self._values]
+
+
+async def test_no_observations_returns_none() -> None:
+    assert await _digraph_centroid(_FakeRepo([]), "id") is None
+
+
+async def test_single_session_centroid_is_that_hash() -> None:
+    out = await _digraph_centroid(_FakeRepo([0xDEADBEEFCAFEF00D]), "id")
+    assert from_bytes8(out) == 0xDEADBEEFCAFEF00D
+
+
+async def test_majority_wins_per_bit() -> None:
+    # 2 of 3 sessions all-ones → every bit majority-set → all ones.
+    out = await _digraph_centroid(_FakeRepo([_ALL_ONES, _ALL_ONES, 0]), "id")
+    assert from_bytes8(out) == _ALL_ONES
+
+
+async def test_tie_is_not_set() -> None:
+    # 1-1 tie per bit: majority requires strictly more than half → 0.
+    out = await _digraph_centroid(_FakeRepo([_ALL_ONES, 0]), "id")
+    assert from_bytes8(out) == 0
+
+
+async def test_garbage_values_skipped() -> None:
+    repo = _FakeRepo([])
+    repo._values = ["not-hex-zz", "deadbeef", to_bytes8(_ALL_ONES).hex()]  # only the last is valid
+    out = await _digraph_centroid(repo, "id")
+    assert from_bytes8(out) == _ALL_ONES
diff --git a/tests/db/test_base_repo.py b/tests/db/test_base_repo.py
index 92825ca7..11f682b9 100644
--- a/tests/db/test_base_repo.py
+++ b/tests/db/test_base_repo.py
@@ -96,8 +96,8 @@ class DummyRepo(BaseRepository):
     async def set_attacker_identity_id(self, a, i): await super().set_attacker_identity_id(a, i)
     async def list_all_identities(self): await super().list_all_identities(); return []
     async def update_identity_merged_into(self, u, w): await super().update_identity_merged_into(u, w)
-    async def update_identity_fingerprints(self, u, *, ja3_hashes=None, hassh_hashes=None, tls_cert_sha256=None):
-        await super().update_identity_fingerprints(u, ja3_hashes=ja3_hashes, hassh_hashes=hassh_hashes, tls_cert_sha256=tls_cert_sha256)
+    async def update_identity_fingerprints(self, u, *, ja3_hashes=None, hassh_hashes=None, tls_cert_sha256=None, kd_digraph_simhash=None):
+        await super().update_identity_fingerprints(u, ja3_hashes=ja3_hashes, hassh_hashes=hassh_hashes, tls_cert_sha256=tls_cert_sha256, kd_digraph_simhash=kd_digraph_simhash)
     # Campaign clustering (this PR)
     async def get_campaign_by_uuid(self, u): await super().get_campaign_by_uuid(u)
     async def list_campaigns(self, limit=50, offset=0): await super().list_campaigns(limit, offset); return []
@@ -260,7 +260,7 @@ async def test_base_repo_coverage():
     await dr.list_all_identities()
     await dr.update_identity_merged_into("a", "b")
     await dr.update_identity_merged_into("a", None)
-    await dr.update_identity_fingerprints("a", ja3_hashes='["x"]', hassh_hashes=None, tls_cert_sha256='["y"]')
+    await dr.update_identity_fingerprints("a", ja3_hashes='["x"]', hassh_hashes=None, tls_cert_sha256='["y"]', kd_digraph_simhash=b"\x00" * 8)
     await dr.get_campaign_by_uuid("a")
     await dr.list_campaigns()
     await dr.count_campaigns()
diff --git a/tests/db/test_kd_digraph_rollup.py b/tests/db/test_kd_digraph_rollup.py
new file mode 100644
index 00000000..43034a91
--- /dev/null
+++ b/tests/db/test_kd_digraph_rollup.py
@@ -0,0 +1,45 @@
+# SPDX-License-Identifier: AGPL-3.0-or-later
+"""kd_digraph_simhash round-trips through update_identity_fingerprints
+and the campaign-clustering projection."""
+from __future__ import annotations
+
+from datetime import datetime, timezone
+
+import pytest
+
+from decnet.web.db.factory import get_repository
+
+
+@pytest.fixture
+async def repo(tmp_path):
+    r = get_repository(db_path=str(tmp_path / "kd.db"))
+    await r.initialize()
+    return r
+
+
+@pytest.mark.asyncio
+async def test_fingerprint_write_and_clustering_read(repo):
+    now = datetime.now(timezone.utc)
+    await repo.create_attacker_identity({
+        "uuid": "id-kd", "first_seen_at": now, "last_seen_at": now,
+    })
+    raw = bytes.fromhex("deadbeefcafef00d")
+    await repo.update_identity_fingerprints("id-kd", kd_digraph_simhash=raw)
+
+    rows = await repo.list_identities_for_clustering()
+    row = next(r for r in rows if r["uuid"] == "id-kd")
+    assert bytes(row["kd_digraph_simhash"]) == raw
+
+
+@pytest.mark.asyncio
+async def test_fingerprint_overwrite_to_none(repo):
+    now = datetime.now(timezone.utc)
+    await repo.create_attacker_identity({
+        "uuid": "id-kd2", "first_seen_at": now, "last_seen_at": now,
+    })
+    await repo.update_identity_fingerprints("id-kd2", kd_digraph_simhash=b"\x01" * 8)
+    # A later pass with no biometric clears it (full-overwrite contract).
+    await repo.update_identity_fingerprints("id-kd2", kd_digraph_simhash=None)
+    rows = await repo.list_identities_for_clustering()
+    row = next(r for r in rows if r["uuid"] == "id-kd2")
+    assert row["kd_digraph_simhash"] is None