scantron
scan machines for unexpected processes, ports, or permissions
purpose
Scantron is a tool for performing a census of processes, ports, protocols, and
file permissions on VMs. Scans can be performed against single VMs, or against
all VMs in a bosh deployment. The intention is to provide a point-in-time scan
which can be analysed offline. Scans are stored in a SQLite database. The CLI
provides a summary report format or the SQLite db can be queried directly.
usage
Whether you scan a single host (direct-scan
) or all VMs in a bosh deployment
(bosh-scan
) the results of the scan will be stored in a SQLite file, or
appended to an existing file.
single host scan
You can perform a direct scan (a scan of a single host) by using the following
command:
scantron direct-scan \
--address scanme.example.com
--username ubuntu \
--password hunter2 \
[--private-key ~/.ssh/id_rsa_scantron]
The password is always required because we use it to sudo
on the machine for
the scan. You may optionally pass a private key for authenticating SSH.
bosh deployment scan
Scantron is typically used in CI jobs and by other machines and so only
supports authenticating with client credentials with a BOSH director at the
moment. You can create a client for use with Scantron like so:
uaac target <bosh uaa host>:<bosh uaa port>
uaac token owner get login admin
uaac client add scantron -s <scantron secret> --authorized_grant_types client_credentials --scope bosh.admin --authorities bosh.admin --access_token_validity 600 --refresh_token_validity 86400
You can then scan a BOSH deployment with the following command:
scantron bosh-scan \
--director-url <bosh address> \
--bosh-deployment <bosh deployment name> \
--client scantron \
--client-secret <scantron secret> \
[--ca-cert bosh.pem]
Multiple deployments can be specified and the results merged into a single database.
scantron bosh-scan \
--director-url <bosh address> \
--bosh-deployment <bosh deployment name> \
--bosh-deployment <second bosh deployment name> \
--client scantron \
--client-secret <scantron secret> \
[--ca-cert bosh.pem]
Note: The scan expects to be able to reach the BOSH machines directly at
the moment so that it can scan the endpoints for their TLS configuration. A
jumpbox is normally a good machine to run this from.
file content check
The file scan can optionally flag files if the content matches a specified regex. For performance optimization
an optional path regex and maximum file size can be specified to limit which files have to be read. The maximum
file size defaults to 1MB.
scantron bosh-scan|direct-scan \
--content <content regex> \
[--path <file path regex>] \
[--max <file size in bytes>]
Regexes use the golang syntax.
checking reports
After you run a scan a report is saved to a SQLite database, by default
database.db
.
With this report it is possible to do the following:
-
Generate a summary of the findings.
scantron report
The report has sections for:
- Externally-accessible processes running as root
- Excluding sshd and rpcbind
- Processes using non-approved SSL/TLS settings
- World-readable files
- Filtered for files from bosh releases (/var/vcap/data/jobs/%)
- Duplicate SSH keys
-
Check to see if any unexpected processes or ports are present in your
cluster.
scantron audit --manifest manifest-of-expected-things.yml
The output from audit
lists the audited host(s) along with either err
or
ok
. Where there are discrepancies with the manifest are highlighted. If
there are any discrepancies the exit code will be 3
, otherwise it is 0
.
Note: Some hand-tweaking may be necessary if there are non-deterministic
ports in your cluster as the generated manifest will contain exactly those
found in the latest scan.
notes
scan filter
Scantron only scans regular files and skips the following directories:
database schema
Scantron produces a SQLite database for scan reports. The database schema can
be found in schema.go.
Scantron does not currently support database migrations. You will be prompted
to create a new database when there are backwards-incompatible changes to the
schema.
Each scan creates a report with many hosts in it. Hosts represent scanned VMs
which contain the list of world writable files and processes running on that
machine. Each process is referenced by the port it is listening on and its
environment variables. TLS information is provided for a port when the port is
expecting TLS connections.
example queries
Finding all of the hosts which are listening on a particular port:
SELECT hosts.NAME
FROM ports
JOIN processes
ON ports.process_id = processes.id
JOIN hosts
ON processes.host_id = hosts.id
WHERE ports.number = 6061
AND upper(ports.state) = "LISTEN"
Scantron audits the hosts, processes, and ports in the database against the
user-generated manifest file.
For Ops Manager where VMs can have the same prefix, such as cloud_controller
and cloud_controller_worker, append "-" to the prefixes: "cloud_controller-"
and "cloud_controller_worker-".
Many hosts (especially those which are based of the BOSH stemcell) will start
processes that bind to an ephemeral, random port when they start. To avoid
caring about these ports when we perform an audit you can add ignore_ports: true
to the process. There is an example of this below for the rpc.statd
process.
This is an example of the manifest file:
specs:
- prefix: cloud_controller-
processes:
- command: sshd
user: root
ports:
- 22
- command: rpcbind
user: root
ports:
- 111
- command: metron
user: vcap
ports:
- 6061
- command: consul
user: vcap
ports:
- 8301
- command: nginx
user: root
ports:
- 9022
- command: ruby
user: vcap
ports:
- 33861
- command: rpc.statd
user: root
ignore_ports: true
development
building
- Install dep, the vendor package manager: https://github.com/golang/dep
go get github.com/pivotal-cf/scantron
cd $GOPATH/src/github.com/pivotal-cf/scantron
dep ensure
# Note: will fail with exit code 1 due to doublestar
test files
./scripts/build
testing
./scripts/test
- There is no step 2.