Re: [PATCH v2] Bluetooth: hci_event: Ignore NULL link key

[joeyli <jlee@suse.com>]

Hi Luiz Augusto von Dentz, 

On Tue, Jul 18, 2023 at 10:22:26AM -0700, Luiz Augusto von Dentz wrote:
> Hi Chun-Yi,
> 
> On Mon, Jul 17, 2023 at 8:43 PM Lee, Chun-Yi <joeyli.kernel@gmail.com> wrote:
> >
> > This change is used to relieve CVE-2020-26555. The description of the
> > CVE:
> >
> > Bluetooth legacy BR/EDR PIN code pairing in Bluetooth Core Specification
> > 1.0B through 5.2 may permit an unauthenticated nearby device to spoof
> > the BD_ADDR of the peer device to complete pairing without knowledge
> > of the PIN. [1]
> 
> Btw, it is probably worth mentioning that in BR/EDR the key generation
> is actually handled in the controller, below HCI.
>

Yes, the key generation be handled by link manager. I will mention it
in patch description. 
 
> > The detail of this attack is in IEEE paper:
> > BlueMirror: Reflections on Bluetooth Pairing and Provisioning Protocols
> > [2]
> >
> > It's a reflection attack. Base on the paper, attacker can induce the
> > attacked target to generate null link key (zero key) without PIN code.
> >
> > We can ignore null link key in the handler of "Link Key Notification
> > event" to relieve the attack. A similar implementation also shows in
> > btstack project. [3]
> 
> Perhaps we could clarify this statement by stating that if we ignore
> the link key it means the stack will not consider the device is bonded
> and will not persist the link key, that said the controller will still
> consider it as paired, so I perhaps we should go one step forward and
> disconnect if we detect such a key is being used.
>

I am new on bluetooth field. Did you mean like this patch? Sending
HCI_Disconnect when we found zero link key?

diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
index ff0c331f53d6..3482031cbbb8 100644
--- a/net/bluetooth/hci_event.c
+++ b/net/bluetooth/hci_event.c
@@ -4698,6 +4700,15 @@ static void hci_link_key_notify_evt(struct hci_dev *hdev, void *data,
        if (!conn)
                goto unlock;
 
+       /* Ignore NULL link key against CVE-2020-26555 */
+       if (!memcmp(ev->link_key, ZERO_KEY, HCI_LINK_KEY_SIZE)) {
+               bt_dev_dbg(hdev, "Ignore NULL link key (ZERO KEY) for %pMR", &ev->bdaddr);
+               hci_disconnect(conn, HCI_ERROR_AUTH_FAILURE);
+               hci_conn_drop(conn);
+               goto unlock;
+       }
+
        hci_conn_hold(conn);
        conn->disc_timeout = HCI_DISCONN_TIMEOUT;
        hci_conn_drop(conn);


Is there anything I'm missing? Thanks a lot!

> > v2:
> > - Used Link: tag instead of Closes:
> > - Used bt_dev_dbg instead of BT_DBG
> > - Added Fixes: tag
> >
> > Fixes: 55ed8ca10f35 ("Bluetooth: Implement link key handling for the management interface")
> > Link: https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2020-26555 [1]
> > Link: https://ieeexplore.ieee.org/abstract/document/9474325/authors#authors [2]
> > Link: https://github.com/bluekitchen/btstack/blob/master/src/hci.c#L3722 [3]
> > Signed-off-by: "Lee, Chun-Yi" <jlee@suse.com>
> > ---
> >  net/bluetooth/hci_event.c | 6 ++++++
> >  1 file changed, 6 insertions(+)
> >
> > diff --git a/net/bluetooth/hci_event.c b/net/bluetooth/hci_event.c
> > index 95816a938cea..ff0c331f53d6 100644
> > --- a/net/bluetooth/hci_event.c
> > +++ b/net/bluetooth/hci_event.c
> > @@ -4684,6 +4684,12 @@ static void hci_link_key_notify_evt(struct hci_dev *hdev, void *data,
> >         bool persistent;
> >         u8 pin_len = 0;
> >
> > +       /* Ignore NULL link key against CVE-2020-26555 */
> > +       if (!memcmp(ev->link_key, ZERO_KEY, HCI_LINK_KEY_SIZE)) {
> > +               bt_dev_dbg(hdev, "Ignore NULL link key (ZERO KEY) for %pMR", &ev->bdaddr);
> > +               return;
> > +       }
> > +
> >         bt_dev_dbg(hdev, "");
> >
> >         hci_dev_lock(hdev);
> > --
> > 2.35.3
> >

Thanks a lot!
Joey Lee