Setting up OSSEC alerts¶
OSSEC is an open source host-based intrusion detection system (IDS) that we use to perform log analysis, file integrity checking, policy monitoring, rootkit detection and real-time alerting. It is installed on the Monitor Server and constitutes that machine’s main function. OSSEC works in a server-agent scheme, that is, the OSSEC server extends its existing functions to the App Server through an agent installed on that server, covering monitoring for both machines.
In order to receive email alerts from OSSEC, you need to supply several settings to Ansible in the playbook for your environment. If you don’t already have a working mail server or don’t know what to do, then see the section below about using Gmail as a fallback option. We assume that you’re working out of the ‘securedrop’ directory you cloned the code into, and editing install_files/ansible-base/prod-specific.yml prior to installing SecureDrop.
What you need:
- The GPG key that OSSEC will encrypt alerts to
- The email address that will receive alerts from OSSEC
- Information for your SMTP server or relay (hostname, port)
- Credentials for the email address that OSSEC will send alerts from
Receiving email alerts from OSSEC requires that you have an SMTP relay to route the emails. You can use an SMTP relay hosted internally, if one is available to you, or you can use a third-party SMTP relay such as Gmail. The SMTP relay does not have to be on the same domain as the destination email address, i.e. smtp.gmail.com can be the SMTP relay and the destination address can be email@example.com.
While there are risks involved with receiving these alerts, such as information leakage through metadata, we feel the benefit of knowing how the SecureDrop servers are functioning is worth it. If a third-party SMTP relay is used, that relay will be able to learn information such as the IP address the alerts were sent from, the subject of the alerts, and the destination email address the alerts were sent to. Only the body of an alert email is encrypted with the recipient’s GPG key. A third-party SMTP relay could also prevent you from receiving any or specific alerts.
The SMTP relay that you use should support SASL authentication and SMTP TLS protocols TLSv1.2, TLSv1.1, and TLSv1. Most enterprise email solutions should be able to meet those requirements.
These are the values you must specify in
- GPG public key used to encrypt OSSEC alerts:
- Fingerprint of key used when encrypting OSSEC alerts:
- The email address that will receive alerts from OSSEC:
- The reachable hostname of your SMTP relay:
- The secure SMTP port of your SMTP relay:
smtp_relay_port(typically 25, 587, or 465. must support TLS encryption)
- Email username to authenticate to the SMTP relay:
- Domain name of the email used to send OSSEC alerts:
- Password of the email used to send OSSEC alerts:
If you don’t know what value to enter for one of these, please ask your
organization’s email administrator for the full configuration before
proceeding. It is better to get these right the first time rather than
changing them after SecureDrop is installed. If you’re not sure of the
smtp_relay_port number, you can use a simple mail client
such as Thunderbird to test different settings or a port scanning tool
such as nmap to see what’s open. You could also use telnet to make sure
you can connect to an SMTP server, which will always transmit a reply
code of 220 meaning “Service ready” upon a successful connection.
smtp_relay mail server hostname is often, but not always,
different from the
sasl_domain, e.g. smtp.gmail.com and gmail.com.
In some cases, authentication or transport encryption mechanisms will vary and you may require later edits to the Postfix configuration (mainly /etc/postfix/main.cf) on the Monitor Server in order to get alerts to work. You can consult Postfix’s official documentation for help, although we’ve described some common scenarios in the troubleshooting section of this document.
If you have your GPG public key handy, copy it to
install_files/ansible-base and then specify the filename in the
ossec_alert_gpg_public_key line of prod-specific.yml.
If you don’t have your GPG key ready, you can run GnuPG on the command line in
order to find, import, and export your public key. It’s best to copy the key
from a trusted and verified source, but you can also request it from keyservers
using the known fingerprint. Looking it up by email address or a shorter key ID
format could cause you to obtain a wrong, malicious, or expired key. Instead, we
recommend you type out your fingerprint in groups of four (just like GPG prints
it) enclosed by double quotes. The reason we suggest this formatting for the
fingerprint is simply because it’s easiest to type and verify correctly. In the
code below simply replace
<fingerprint> with your full, space-separated
Download your key and import it into the local keyring:
gpg --recv-key "<fingerprint>"
It is important you type this out correctly. If you are not copy-pasting this command, we recommend you double-check you have entered it correctly before pressing enter.
Again, when passing the full public key fingerprint to the
--recv-key command, GPG
will implicitly verify that the fingerprint of the key received matches the
If GPG warns you that the fingerprint of the key received does not match the one requested do not proceed with the installation. If this happens, please email us at firstname.lastname@example.org.
Next we export the key to a local file.
gpg --export -a "<fingerprint>" > ossec.pub
Copy the key to a directory where it’s accessible by the SecureDrop installation:
cp ossec.pub install_files/ansible-base/
The fingerprint is a unique identifier for an encryption (public) key. The
short and long key ids correspond to the last 8 and 16 hexadecimal digits of the
fingerprint, respectively, and are thus a subset of the fingerprint. The value
ossec_gpg_fpr must be the full 40 hexadecimal digit GPG fingerprint for
this same key, with all capital letters and no spaces. The following command
will retrieve and format the fingerprint per our requirements:
gpg --with-colons --fingerprint "<fingerprint>" | grep "^fpr" | cut -d: -f10
Next you specify the e-mail that you’ll be sending alerts to, as
ossec_alert_email. This could be your work email, or an alias for a
group of IT administrators at your organization. It helps for your mail
client to have the ability to filter the numerous messages from OSSEC
into a separate folder.
Now you can move on to the SMTP and SASL settings, which are
straightforward. These correspond to the outgoing e-mail address used to
send the alerts instead of where you’re receiving them. If that e-mail
is email@example.com, the
sasl_username would be ossec and
sasl_domain would be news-org.com.
The Postfix configuration enforces certificate verification, and
requires both a valid certificate and STARTTLS support on the SMTP
relay. By default the system CAs will be used for validating the relay
certificate. If you need to provide a custom CA to perform the
validation, copy the cert file to
install_files/ansible-base add a
new variable to
MyOrg.crt is the filename. The file will be copied to the
/etc/ssl/certs_local and the system CAs will be ignored
when validating the SMTP relay TLS certificate.
prod-specific.yml, exit the editor and proceed with the
installation by running the playbooks.
Using Gmail for OSSEC alerts¶
It’s easy to get SecureDrop to use Google’s servers to deliver the alerts, but it’s not ideal from a security perspective. This option should be regarded as a backup plan. Keep in mind that you’re leaking metadata about the timing of alerts to a third party — the alerts are encrypted and only readable to you, however that timing may prove useful to an attacker.
First you should sign up for a new
account. While it’s
technically possible to use an existing Gmail account, it’s best to
compartmentalize these alerts from any of your other activities. Choose
a strong and random passphrase for the new account. Skip the creation of
a Google+ profile and continue straight to Gmail. Next, enable Google’s
2-Step Verification. With
2-Step Verification enabled, you won’t use the normal account password
in this configuration — it will not work; instead you must navigate
(using the settings in the top right) to Account > Signing in > App
passwords, and generate a new App password which you will use as the
Once the account is created you can log out and provide the values for
sasl_username as your new Gmail username (without the domain),
sasl_domain, which is typically gmail.com (or your custom Google
Apps domain), and
sasl_passwd. Remember to use the App password
generated from the 2-step config for
sasl_passwd, as the primary
account password won’t work. The
smtp_relay is smtp.gmail.com and
smtp_relay_port is 587.
Configuring fingerprint verification¶
If you run your own mail server, you may wish to increase the security
level used by Postfix for sending mail to
fingerprint, rather than
secure. Doing so will require an exact match for the fingerprint of
TLS certificate on the SMTP relay. The advantage to fingerprint
verification is additional security, but the disadvantage is potential
maintenance cost if the fingerprint changes often. If you manage the
mail server and handle the certificate rotation, you should update the
SecureDrop configuration whenever the certificate changes, so that OSSEC
alerts continue to send. Using fingerprint verification does not work
well for popular mail relays such as smtp.gmail.com, as those
fingerprints can change frequently, due to load balancing or other
You can retrieve the fingerprint of your SMTP relay by running the
command below (all on one line). Please note that you will need to
587 with the correct domain and port
for your SMTP relay.
openssl s_client -connect smtp.gmail.com:587 -starttls smtp < /dev/null 2>/dev/null | openssl x509 -fingerprint -noout -in /dev/stdin | cut -d'=' -f2
If you are using Tails, you will not be able to connect directly with
openssl s_client due to the default firewall rules. To get around
this, proxy the requests over Tor by adding
torify at the beginning
of the command. The output of the command above should look like the
Finally, add a new variable to
smtp_relay_fingerprint, like so:
Specifying the fingerprint will configure Postfix to use it for
verification on the next playbook run. (To disable fingerprint
verification, simply delete the variable line you added, and rerun the
prod-specific.yml, exit the editor and proceed
with the installation by running the
Some OSSEC alerts should begin to arrive as soon as the installation has finished.
The easiest way to test that OSSEC is working is to SSH to the Monitor
Server and run
service ossec restart. This will trigger an Alert
level 3 saying: “Ossec server started.”
So you’ve finished installing SecureDrop, but you haven’t received any OSSEC alerts. First, check your spam/junk folder. If they’re not in there, then most likely there is a problem with the email configuration. In order to find out what’s wrong, you’ll have to SSH to the Monitor Server and take a look at the logs. To examine the mail log created by Postfix, run the following command:
The output will show you attempts to send the alerts and provide hints as to what went wrong. Here’s a few possibilities and how to fix them:
|Connection timed out||
Check that the hostname and port is correct in the relayhost line of
|Server certificate not verified||
Check that the relay certificate is valid (for more detailed help, see Troubleshooting SMTP TLS). Consider adding
to update when finished.
After making changes to the Postfix configuration, you should run
service postfix reload and test the new settings by restarting the
If you change the SMTP relay port after installation for any
reason, you must update the
smtp_relay_port variable in the
prod-specific.yml file, then rerun the Ansible playbook.
As a general best practice, we recommend modifying and
rerunning the Ansible playbook instead of manually editing
the files live on the servers, since values like
are used in several locations throughout the config.
Useful log files for OSSEC¶
Other log files that may contain useful information:
- Includes lines for sending mail containing OSSEC alerts.
- Messages related to grsecurity, AppArmor and iptables.
- OSSEC’s general operation is covered here.
- Contains details of every recent OSSEC alert.
Remember to encrypt any log files before sending via email, for example to firstname.lastname@example.org, in order to protect security-related information about your organization’s SecureDrop instance.
Troubleshooting SMTP TLS¶
Your choice of SMTP relay server must support STARTTLS and have a valid
server certificate. By default, the Monitor Server’s Postfix
configuration will try to validate the server certificate using the
default root store (in Ubuntu, this is maintained in the
ca-certificates package). You can override this by setting
smtp_relay_cert_override_file as described earlier in this document.
In either situation, it can be helpful to use the
line tool to verify that you can successfully connect to your chosen
SMTP relay securely. We recommend doing this before running the
playbook, but it can also be useful as part of troubleshooting OSSEC
email send failures.
In either case, start by attempting to make a STARTTLS connection to
smtp_relay:smtp_relay_port (get the values from your
prod-specific.yml file). On a machine running Ubuntu, run the
openssl command, replacing
smtp_relay_port with your specific values:
openssl s_client -showcerts -starttls smtp -connect smtp_relay:smtp_relay_port < /dev/null 2> /dev/null
Note that you will not be able to run this command on the Application Server because of the firewall rules. You can run it on the Monitor Server, but you will need to run it as the Postfix user (again, due to the firewall rules):
sudo -u postfix openssl s_client -showcerts -starttls smtp -connect smtp.gmail.com:587 < /dev/null 2> /dev/null
If the command fails with “Could not connect” or a similar message, then
this mail server does not support STARTTLS. Verify that the values you
are using for
smtp_relay_port are correct. If
they are, you should contact the admin of that relay and talk to them
about supporting STARTTLS, or consider using another relay that already
If the command succeeds, the first line of the output should be
“CONNECTED” followed by a lot of diagnostic information about the
connection. You should look for the line that starts with “Verify return
code”, which is usually one of the last lines of the output. Since we
did not give
openssl any information about how to verify
certificates in the previous command, it should be a non-zero value
(indicating verification failed). In my case, it is
Verify return code: 20 (unable to get local issuer certificate),
which indicates that openssl does not know how to build the certificate
chain to a trusted root.
If you are using the default verification setup, you can check whether
your cert is verifiable by the default root store with
openssl s_client -CApath /etc/ssl/certs -showcerts -starttls smtp -connect smtp_relay:smtp_relay_port < /dev/null 2> /dev/null
For example, if I’m testing Gmail as my SMTP relay
smtp.gmail.com:587), running the
openssl with the default root
store results in
Verify return code: 0 (ok) because their
certificate is valid and signed by one of the roots in the default
store. This indicates that can be successfully used to securely relay
email in the default configuration of the Monitor Server.
If your SMTP relay server does not successfully verify, you should use the return code and its text description to help you diagnose the cause. Your cert may be expired, in which case you should renew it. It may not be signed by a trusted CA, in which case you should obtain a signature from a trusted CA and install it on the mail server. It may not have the right hostnames in the Common Name or Subject Alternative Names, in which case you will need to generate a new CSR with the correct hostnames and then obtain a new certificate and install it. Etc., etc.
If you are not using the the default verification setup, and
intentionally do not want to use a certificate signed by one of the
default CA’s in Ubuntu, you can still use
openssl to test whether
you can successfully negotiate a secure connection. Begin by copying
your certificate file (
prod-specific.yml) to the computer you are using for testing. You
-CAfile to test if your connection will succeed using your
custom root certificate:
openssl s_client -CAfile /path/to/smtp_relay_cert_override_file -showcerts -starttls smtp -connect smtp_relay:smtp_relay_port < /dev/null 2> /dev/null
Finally, if you have a specific server in mind but are not sure what
certificate you need to verify the connection, you can use the output of
openssl s_client to figure it out. Since we have
openssl prints the entire certificate chain it receives
from the server. A properly configured server will provide all of the
certificates in the chain up to the root cert, which needs to be
identified as “trusted” for the verification to succeed. To see the
chain, find the part of the output that start with
Certificate chain. It will look something like this (example from
smtp.gmail.com, with certificate contents snipped for brevity):
--- Certificate chain 0 s:/C=US/ST=California/L=Mountain View/O=Google Inc/CN=smtp.gmail.com i:/C=US/O=Google Inc/CN=Google Internet Authority G2 -----BEGIN CERTIFICATE----- <snip> -----END CERTIFICATE----- 1 s:/C=US/O=Google Inc/CN=Google Internet Authority G2 i:/C=US/O=GeoTrust Inc./CN=GeoTrust Global CA -----BEGIN CERTIFICATE----- <snip> -----END CERTIFICATE----- 2 s:/C=US/O=GeoTrust Inc./CN=GeoTrust Global CA i:/C=US/O=Equifax/OU=Equifax Secure Certificate Authority -----BEGIN CERTIFICATE----- <snip> -----END CERTIFICATE----- ---
The certificates are in reverse order from leaf to root.
handily prints the Subject (
s:) and Issuer (
i:) information for
each cert. In order to find the root certificate, look at the Issuer of
the last certificate. In this case, that’s
Equifax Secure Certificate Authority. This is the root certificate
that issued the first certificate in the chain, and it is what you need
to tell Postfix to use in order to trust the whole connection.
Actually obtaining this certificate and establishing trust in it is beyond the scope of this document. Typically, if you are using your own SMTP relay with a custom CA, you will be able to obtain this certificate from an intranet portal or someone on your IT staff. For a well-known global CA, you can obtain it from the CA’s website. For example, a quick search for “Equifax Secure Certificate Authority” finds the web page of GeoTrust’s Root Certificates, which have accompanying background information and are available for download.
Once you have the root certificate file, you can use
-CAfile to test
that it will successfully verify the connection.
Analyzing the Alerts¶
Understanding the contents of the OSSEC alerts requires a background and knowledge in Linux systems administration. They may be confusing, and at first it will be hard to tell between a genuine problem and a fluke. You should examine these alerts regularly to ensure that the SecureDrop environment has not been compromised in any way, and follow up on any particularly concerning messages with direct investigation.
Common OSSEC Alerts¶
The SecureDrop Application and Monitor Servers reboot every night, as part of the unattended upgrades process. When the servers come back up, OSSEC will start again and report the change in status. Therefore you should receive an email alert every morning containing text similar to:
Received From: mon->ossec-monitord Rule: 502 fired (level 3) -> "Ossec server started." Portion of the log(s): ossec: Ossec started.
This is a normal alert, and informs you that the system is working as expected.
Similarly, your SecureDrop Application and Monitoring Servers will
check for application updates on your servers. Should your servers require
updates, OSSEC will alarm because the packages binaries will have changed
Below is a sample alert, but you may see any number of these records in the
logs. This will happen in batches so these emails might be longer than the
below alert. You should also see them in an email named
File Changes. To verify this activity matches the package history, you
can review the logs in
Received From: (app) Rule: 2902 fired (level 7) -> "New (Debian Package) installed." Portion of the log(s): status installed <package name> <version>
This is a normal alert, it tells you your system is up-to-date and patched.
Occassionally your SecureDrop Servers will send an alert for failing to connect to Tor relays. Since SecureDrop runs as a Tor Onion Service, it is possible for Tor connections to timeout or become overloaded.
Received From: (app) Rule: 1002 fired (level 2) -> "Unknown problem somewhere in the system." Portion of the log(s): [warn] Your Guard <name> ($fingerprint) is failing a very large amount of circuits. Most likely this means the Tor network is overloaded, but it could also mean an attack against you or potentially the guard itself.