Intrusion Detection #
Let’s imagine that the client application generates requests and sends them through the chain of Acra’s services to receive a decrypted plaintext in the end.
What if the attackers could compromise the app in such a way that they are able to alter its behaviour? In this case, a various threats could come from a compromised application (SQL injection, remote/local code execution, etc.). Our idea is that since AcraServer is the only entity that decrypts sensitive data and sees all the requests addressed to it, it is AcraServer’s responsibility to distinguish good behaviour from bad one and to perform appropriate action.
For intrusion detection, Acra uses poison records, also known as honey tokens (or canarytokens, or trip wires).
We call them poison records because they look like ordinary encrypted records but contain poison.
Poison records #
Poison records are specifically designed to sit quietly in the database; they are never queried by legitimate users under normal circumstances. They look like any other encrypted records, and it’s impossible to distinguish them from
Technically speaking, poison records are encrypted data (binary, strings, int or whatever suits your database design), placed in particular tables/columns/cells.
However, you should place poison records carefully – put them into columns/tables/calls that should NOT be read by your application under normal circumstances. For example, create a dummy user record “Sam Smith” and put poison records into its row.
In this case, poison records will only be included in the outputs of requests from malicious applications that read more data than they should, i.e. using
SELECT * requests. The goal of using poison records is simple — to detect adversaries trying to download full tables/database from the application server or trying to run full scans in their injected queries.
When AcraServer or AcraTranslator find that encrypted blob is not a “normal” AcraStruct/AcraBlock, but a poison record, they will inform your Ops team of untypical behaviour and can block suspicious requests.
For more technical details you can check out other blog posts related to poison records:
- Explain Like I’m Five: Poison Records (Honeypots for Database Tables)
- Poison Records In Acra – Database Honeypots For Intrusion Detection
Command line flags #
AcraServer has a set of flags for more precise work configuration with Poison records. Same flags are available for AcraTranslator.
Turn on poison record detection. If shutdown is not enabled, AcraServer just logs the poison record detection and returns decrypted data. Default is
On detecting poison record: log about poison record detection, stop and shutdown. Default is
On detecting poison record: log about poison record detection, execute script, return decrypted data.
Usage example #
AcraServer’s key storage contains a special key, which is used for recognition of poison records. This key is generated either with
acra-keys, or upon running poison record generation utility
acra-pisonrecordmaker (but you will have to move the keys into AcraServer’s key storage manually).
Before running any commands, make sure you have installed
acra-poisonrecordmaker utility and
ACRA_MASTER_KEY stored as an environmental variable.
If you haven’t generated keys before, checkout out key generation instructions. After generating
ACRA_MASTER_KEY, assign it to a variable like this:
export ACRA_MASTER_KEY=$(echo -n "My_Very_Long_Key_Phrase_ge_32_chars" | base64)
acra-poisonrecordmaker binary command:
Your output will be a base64-encoded poison record that you should decode and insert into your database as binary data. It will look something like this:
acra-poisonrecordmakerkey folder must have proper permissions (as set originally by acra-keymaker):
- folder 700
- private keys 600
The Intrusion detection with Acra example project illustrates how to use intrusion detection functionality of Acra data protection suite.