appguard-server

AppGuard is an application-level firewall for general applications.

Currently, AppGuard client libraries for Express and SMTP servers are available, and support for more applications is planned for the future.

Methodology

AppGuard consists of a Rust-based gRPC server that receives and handles layer 7 network traffic from its clients.

The server is responsible for:

• logging layer 7 network traffic
  • in console
  • in an SQLite database
• gathering additional IP information about each TCP connection (e.g., geolocation, ASN, organization, blacklist status), via
  • the ipapi.co API
  • MaxMind databases from ipinfo.io updated daily
  • the ipsum daily updated feed of malicious IP addresses
• matching the traffic against a set of user-defined firewall rules that can be dynamically updated at runtime
• returning the result of the firewall rule matching to the client library

The client libraries are responsible for:

• sending the layer 7 network traffic to the gRPC server
• receiving the result of the firewall rule matching from the gRPC server, and acting accordingly

The contract between the server and the client libraries is defined in the appguard-protobuf/appguard.proto file, which adheres to the Protocol Buffers specification.

Firewall specification

The firewall is defined in a JSON file, which is loaded by the server at startup and can be updated at runtime.

A firewall consists of a collection of expressions.
Each expression is associated with a policy (either allow or deny) and a set of tokens.
The expression's policy is applied if the tokens match the incoming traffic.

Firewall tokens are specified as a mathematical expression including parenthesis, operators (AND, OR), and predicates.
Predicates include a matching condition (e.g., equal, not_equal, lower_than, contains, etc.) and a value.
The value of a predicate explicitly refers to a field of the analyzed traffic (e.g., the user agent, the source IP, a particular header, the response code, etc.).

For an example of a firewall specification, see the firewall.json file.

Internally, each mathematical expression of the firewall is parsed and converted to its equivalent Reverse Polish Notation (RPN), also known as postfix form.
The RPN form is then used to more efficiently evaluate the expression against the incoming traffic.