NFD » ndn-cxx

# Validator Configuration File Format

You can set up a `Validator` via a configuration file. 

Next, we will show you how to write a configuration file.

The configuration file consists of **rules** and **trust-anchors** that will be used in validation.

**Rules** tell the validator how to validate a packet, while **trust-anchors** tell the validator which certificates are valid immediately.

Here is an example of configuration file containing two rules.

    rule

    {

      id "Simple Rule"

      for data

      filter

      {

        type name

        name /localhost/example

        relation is-prefix-of

      }

      checker

      {

        type customized

        sig-type rsa-sha256

        key-locator

        {

          type name

          name /ndn/edu/ucla/KEY/yingdi/ksk-1234/ID-CERT

          relation equal

        }

      }

    }

    rule

    {

      id "Testbed Validation Rule"

      for data

      checker

      {

        type hierarchical

        sig-type rsa-sha256

      }

    }

    trust-anchor

    {

      type file

      file-name "testbed-trust-anchor.cert"

    }

* <font color='red'>**ATTENTION: The order of rules MATTERS!**</font>

A rule can be broken into two parts: 

* The first part is to qualify packets to which the rule can be applied;

* The second part is to check whether further validation process is necessary.

When receiving a packet, the validator will apply rules in the configuration file one-by-one against the packet,

until finding a rule that the packet qualifies for.

And the second part of the matched rule will be used to check the validity of the packet.

If the packet cannot qualify for any rules, it is treated as an invalid packet.

Once a packet has been matched by a rule, the rest rules will not be applied against the packet.

Therefore, you should always put the most specific rule to the top, otherwise it will become useless.

In the example configuration,

the first rule indicates that all the data packets under the name prefix "/localhost/example" must be signed by a key whose certificate name is "/ndn/edu/ucla/KEY/yingdi/ksk-1234/ID-CERT".

If a packet does not have a name under prefix "/localhost/example", validator will skip the first rule and apply the second rule.

The second rule indicates that any data packets must be validated along a hierarchy. 

And a certificate stored in a file "testbed-trust-anchor.cert" is valid.

## Rules in general

A rule has four types of properties: **id**, **for**, **filter**, and **checker**.

The property **id** uniquely identifies the rule in the configuration file.

As long as being unique, any name can be given to a rule, e.g., "Simple Rule", "Testbed Validation Rule".

A rule must have one and only one **id** property.

A rule is either used to validate an interest packet or a data packet. 

This information is specified in the property **for**.

Only two value can be specified: **data** and **interest**.

A rule must have one and only one **for** property.

* <font color='red'>**ATTENTION: For [signed interest](http://redmine.named-data.net/projects/nfd/wiki/Signed_Interests), rules should cover the prefix before the ``timestamp``**</font>.

The property **filter** further constrains the packets that can be checked by the rule.

Filter property is not required in a rule, in this case, the rule will capture all the packets passed to it.

A rule may contain more than one filters, in this case, a packet can be checked by a rule only if the packet satisfies all the filters.

* <font color='red'>**ATTENTION: A packet that satisfies all the filters may not be valid**</font>.

The property **checker** defines the conditions that a matched packet must fulfill to be treated as a valid packet.

A rule must have at least one **checker** property, a packet is treated as invalid if it cannot pass none of the checkers.

**filter** and **checker** have their own properties.

Next, we will introduce them separately.

## Filter Property

Filter has its own **type** property.

Although a rule may contain more than one filters, there is at most one filter of each type.

So far, only one type of filter is defined: **name**.

In other word, only one filter can be specified in a rule for now.

### Name Filter

There are two ways to express the conditions on name. 

The first way is to specify a relationship between the packet name and a particular name.

In this case, two more properties are required: **name** and **relation**.

A packet can fulfill the condition if the **name** has a **relation* to the packet name.

Three types of **relation** has been defined: **equal**, **is-prefix-of**, **is-strict-prefix-of**. 

For example, a filter 

    filter

    {

      type name

      name /localhost/example

      relation equal

    }

shall only capture a packet with the exact name "/localhost/example".

And a filter

    filter

    {

      type name

      name /localhost/example

      relation is-prefix-of

    }

shall capture a packet with name "/localhost/example" or "/localhost/example/data", but cannot catch a packet with name "/localhost/another_example".

And a filter

    filter

    {

      type name

      name /localhost/example

      relation is-strict-prefix-of

    }

shall capture a packet with name "/localhost/example/data", but cannot catch a packet with name "/localhost/example".

The second way is to specify an [[Regex|NDN Regular Expression]] that can match the packet.

In this case, only one property **regex** is required.

For example, a filter 

    filter

    {

      type name

      regex ^[^<KEY>]*<KEY><>*<ksk-.*><ID-CERT>$

    }

shall capture all the identity certificates. 

## Checker Property

Passing all the filters in a rule only indicates that a packet can be checked using the rule, 

and it does not necessarily implies that the packet is valid.

The validity of a packet is determined by the property **checker**, which defines the conditions that a valid packet must fulfill.

Same as **filter**, **checker** has a property **type**.

We have defined three types of checkers: **customized**, and **hierarchical**, and **fixed-signer**.

As suggested by its name, **customized** checker allows you to customize the conditions according to specific requirements.

**hierarchical** checker and **fixed-signer** checker are pre-defined shortcuts, which specify specific trust models separately.

### Customized Checker

So far, we only allow three customized properties in a customized checker: **sig-type**, **key-locator**.

All of them are related to the `SignatureInfo` of a packet. 

    checker

    {

      type customized

      sig-type ...

      key-locator

      {

        ...

      }

    }

The property **sig-type** specifies the acceptable signature type.

Right now two signature types have been defined: **rsa-sha256** (which is a strong signature type) and **sha256** (which is a weak signature type).

If sig-type is sha256, then **key-locator** will be ignored.

Validator will simply calculate the digest of a packet and compare it with the one in `SignatureValue`.

If sig-type is rsa-sha256, you have to further customize the checker with **key-locator**.

The property **key-locator** which specifies the conditions on `KeyLocator`.

If the **key-locator** property is specified, it requires the existence of the `KeyLocator` field in `SignatureInfo`.

Although there are more than one types of `KeyLocator` defined in the [Packet Format](http://named-data.net/doc/ndn-tlv/signature.html), 

**key-locator** property only supports one type: **name**:

    key-locator

    {

      type name

      ...

    }

Such a key-locator property specifies the conditions on the certificate name of the signing key.

Since the conditions are about name, they can be specified in the same way as the name filter.

For example, a checker could be:

    checker

    {

      type customized

      sig-type rsa-sha256

      key-locator

      {

        type name

        name /ndn/edu/ucla/KEY/yingdi/ksk-1234/ID-CERT

        relation equal

      }

    }

This checker property requires that the packet must have a rsa-sha256 signature generated by a key whose certificate name is "/ndn/edu/ucla/KEY/yingdi/ksk-1234/ID-CERT".

Besides the two ways to express conditions on the `KeyLocator` name (name and regex), 

you can further constrain the `KeyLocator` name using the information extracted from the packet name.

This third type of condition is expressed via a property **hyper-relation**.

The **hyper-relation** property consists of three parts:

* an NDN regular expression that can extract information from packet name

* an NDN regular expression that can extract information from `KeyLocator` name

* relation from the part extracted from `KeyLocator` name to the one extracted from the packet name

For example, a checker:

    checker

    {

      type customized

      sig-type rsa-sha256

      key-locator

      {

        type name

        hyper-relation

        {

          k-regex ^([^<KEY>]*)<KEY>(<>*)<ksk-.*><ID-CERT>$

          k-expand \\1\\2

          h-relation is-prefix-of

          p-regex ^(<>*)$

          p-expand \\1

        }

      }

    }

requires the packet name must be under the corresponding namespace of the `KeyLocator` name.

In some cases, you can even customize checker with another property 

For example:

    checker

    {

      type customized

      sig-type rsa-sha256

      key-locator

      {

        type name

        hyper-relation

        {

          k-regex ^([^<KEY>]*)<KEY>(<>*)<ksk-.*><ID-CERT>$

          k-expand \\1\\2

          h-relation is-prefix-of

          p-regex ^(<>*)$

          p-expand \\1

        }

      }

    }

### Hierarchical Checker

As implied by its name, hierarchical checker requires that the packet name must be under the namespace of the packet signer.

A hierarchical checker:

    checker

    {

      type hierarchical

      sig-type rsa-sha256

    }

is equivalent to a customized checker:

    checker

    {

      type customized

      sig-type rsa-sha256

      key-locator

      {

        type name

        hyper-relation

        {

          k-regex ^([^<KEY>]*)<KEY>(<>*)<ksk-.*><ID-CERT>$

          k-expand \\1\\2

          h-relation is-prefix-of

          p-regex ^(<>*)$

          p-expand \\1

        }

      }

    }

### Fixed-Signer Checker

In some cases, you only accept packets signed with pre-trusted certificates, i.e. "one-step validation".

Such a trust model can be expressed with **fixed-signer** checker.

And you only need to specify the trusted certificate via property **signer**. 

The definition of **signer** is the same as **trust-anchor**.

For example:

    checker

    {

      type fixed-signer

      sig-type rsa-sha256

      signer

      {

        type file

        file-name "trusted-signer.cert"

      }

      signer

      {

        type base64

        base64-string "Bv0DGwdG...amHFvHIMDw=="

      }

    }

## Trust Anchors

Although **trust-anchor** is always not required in the configuration file (for example, if fixed-signer checker is used), 

it is very common to have a few trust-anchors in the configuration file, otherwise most packets cannot be validated.

A configuration file may contain more than one trust anchors, but the order of trust anchors does not matter.

The structure of trust-anchor is same as the **signer** in fixed-signer checker, for example:

    trust-anchor

    {

      type file

      file-name "trusted-signer.cert"

    }

    trust-anchor

    {

      type base64

      base64-string "Bv0DGwdG...amHFvHIMDw=="

    }

You may also specify a trust anchor directory as following:

    trust-anchor

    {

      type dir

      dir /usr/local/ndn/keys

    }

Validator will load every file as a trust anchor from the directory.

If the content in the directory will change during the runtime, 

you can specify **refresh** property to ask Validator to periodically reload trust anchors. 

For example:

    trust-anchor

    {

      type dir

      dir /usr/local/ndn/keys

      refresh 1h

    }

indicates that validator should reload trust anchors from this directory every hour.

There three options of time units: `h` for hour, `m` for minutes, and `s` for seconds.

If the refresh period is set to `0`, then the default refresh period (`1h`) will be used.

If the refresh property is not specified, then trust anchors will be loaded for only one time.

There is another special trust anchor **any**.

As long as such a trust-anchor is defined in config file, packet validation will be turned off.

<font color=red>**ATTENTION!! Such a type of trust anchor is dangerous. 

You should used it only when you want to disable packet validation temporarily (e.g, debugging code, building a demo).**</font>

    trust-anchor

    {

      type any

    }

## Example Configuration For NLSR

The trust model of NLSR is semi-hierarchical.

An example certificate signing hierarchy is:

                                            root 

                                             |  	    

                              +--------------+---------------+

                            site1                          site2

                              |                              |		   

                    +---------+---------+                    +

                 operator1           operator2            operator3

                    |                   |                    | 

              +-----+-----+        +----+-----+        +-----+-----+--------+

           router1     router2  router3    router4  router5     router6  router7

              |           |        |          |        |           |        |

              +           +        +          +        +           +        +  	      

            NLSR        NSLR     NSLR       NSLR     NSLR        NSLR     NSLR

However, entities name may not follow the signing hierarchy, for example:

Entity   | Identity Name                                     | Example                         | Certificate Name Example

-------- | ------------------------------------------------- | ------------------------------- | ------------------------------------------------

root     | /\<network\>                                      | /ndn                            | /ndn/KEY/ksk-1/ID-CERT/%01

site     | /\<network\>/\<site\>                             | /ndn/edu/ucla                   | /ndn/edu/ucla/KEY/ksk-2/ID-CERT/%01

operator | /\<network\>/\<site\>/%C1.O.N./\<operator-id\>    | /ndn/edu/ucla/%C1.O.N./op1      | /ndn/edu/ucla/%C1.O.N./op1/KEY/ksk-3/ID-CERT/%01

router   | /\<network\>/\<site\>/%C1.O.R./\<router-id\>      | /ndn/edu/ucla/%C1.O.R./rt1      | /ndn/edu/ucla/%C1.O.R./rt1/KEY/ksk-4/ID-CERT/%01

NLSR     | /\<network\>/\<site\>/%C1.O.R./\<router-id\>/NLSR | /ndn/edu/ucla/%C1.O.R./rt1/NLSR | /ndn/edu/ucla/%C1.O.R./rt1/NLSR/KEY/ksk-5/ID-CERT/%01

Assume that a typical NLSR data name is "/ndn/edu/ucla/%C1.O.R./rt1/NLSR/LSA/LSType.1/%01".

Then, the exception of naming hierarchy is "operator-router".

So we can write a configuration file with three rules.

The first one is a customized rule that capture the normal NLSR data.

The second one is a customized rule that handles the exception case of the hierarchy (operator->router).

And the last one is a hierarchical rule that handles the normal cases of the hierarchy.

We put the NLSR data rule to the first place, because NLSR data packets are the most frequently checked.

The hierarchical exception rule is put to the second, because it is more specific than the last one.

And here is the configuration file:

    rule

    {

      id "NSLR LSA Rule"

      for data

      filter

      {

        type name

        regex ^[^<NLSR><LSA>]*<NLSR><LSA>

      }

      checker

      {

        type customized

        sig-type rsa-sha256

        key-locator

        {

          type name

          hyper-relation

          {

            k-regex ^([^<KEY>]*)<KEY><ksk-.*><ID-CERT>$

            k-expand \\1

            h-relation equal

            p-regex ^([^<NLSR><LSA>]*)<NLSR><LSA><LSType\.\d><>$

            p-expand \\1

          }

        }

      }

    }

    rule

    {

      id "NSLR Hierarchy Exception Rule"

      for data

      filter

      {

        type name

        regex ^[^<KEY><%C1.O.R.>]*<%C1.O.R.><><KEY><ksk-.*><ID-CERT><>$

      }

      checker

      {

        type customized

        sig-type rsa-sha256

        key-locator

        {

          type name

          hyper-relation

          {

            k-regex ^([^<KEY><%C1.O.N.>]*)<%C1.O.N.><><KEY><ksk-.*><ID-CERT>$

            k-expand \\1

            h-relation equal

            p-regex ^([^<KEY><%C1.O.R.>]*)<%C1.O.R.><><KEY><ksk-.*><ID-CERT><>$

            p-expand \\1

          }

        }

      }

    }

    rule

    {

      id "NSLR Hierarchical Rule"

      for data

      filter

      {

        type name

        regex ^[^<KEY>]*<KEY><ksk-.*><ID-CERT><>$

      }

      checker

      {

        type hierarchical

        sig-type rsa-sha256

      }

    }

    trust-anchor

    {

      type file

      file-name "testbed-trust-anchor.cert"

    }

## Example Configuration For NRD

Assume NRD allows any valid testbed certificate to register prefix, the configuration file could be written as:

    rule

    {

      id "NRD Prefix Registration Command Rule"

      for interest

      filter

      {

        type name

        regex ^<localhost><nrd>[<register><unregister><advertise><withdraw>]

      }

      checker

      {

        type customized

        sig-type rsa-sha256

        key-locator

        {

          type name

          regex ^[^<KEY>]*<KEY><>*<ksk-.*><ID-CERT>$

        }

      }

    }

    rule

    {

      id "Testbed Hierarchy Rule"

      for data

      filter

      {

        type name

        regex ^[^<KEY>]*<KEY><>*<ksk-.*><ID-CERT><>$

      }

      checker

      {

        type hierarchical

        sig-type rsa-sha256

      }

    }

    trust-anchor

    {

      type file

      file-name "testbed-trust-anchor.cert"

    }