5.1 Rule-Based Classification

The classification process evaluates records with a set of rules to determine it’s categories.

The ASP considers that an object categorized as ${\Gamma }_n$ is therefore ${\Gamma }_n$ compliant and must have satisfied all the rules of ${\Gamma }_n$, i.e.:

• A transaction is categorised as AML compliant because it has satisfied AML rules such as Sender is not in the OFAC list.
• A person is categorised as KYC compliant because it has satisfied KYC rules such as Person has provided a valid ID.
• A vote is categorised as Valid Vote because it has satisfied voting rules such as Vote is casted within the voting period.
• A document is categorised as Approved Document because it has satisfied document rules such as Document is signed & reviewed by an auditor.

1. The Compliance Domain:

Given a compliance predicate $P$ and it’s propositional function: $$P(x,\Gamma )=x\text{ is }\Gamma \text{ compliant }$$

• Where $\Gamma$ is a set of atomic rules: $$\Gamma =r_1,r_2,r_3,\dots ,r_n$$

$\Gamma$ is considered a Compliance Domain only if the following holds true:

$$P(x,\Gamma )⟺\forall r\in \Gamma :r(x)=true$$ $$P(y,\Gamma )⟺\exists r\in \Gamma :r(y)=false$$ $$x\in X,y\in Y,X\subset Z,Y\subset Z$$

Where:

• $x$ is compliant with $\Gamma$ if and only if $x$ satisfies all the rules in $\Gamma$.
• $y$ is not compliant with $\Gamma$ if and only if $y$ does not satisfy all the rules in $\Gamma$.
• $X$ is a set of transactions that are compliant with $\Gamma$.
• $Y$ is a set of transactions that are not compliant with $\Gamma$.
• $X$ and $Y$ are subsets of the universal set $Z$.

Example:

If the statements are true:

• A set of transactions ($X$) exists such that each transaction ($x$) is AML ($\Gamma$) compliant.
• A set of transactions ($Y$) exists such that each transaction ($y$) is not AML ($\Gamma$) compliant.
• $X$ and $Y$ belong to a super set of transactions $Z$.

Then Compliance Domain ($\Gamma$) exists as a set of rules for which holds true for all transactions in $X$ and false for all transactions in $Y$.

2. Defining Compliance Rules:

By identifiying ($\Gamma$) and set $X$ & $Y$ it is then possible to derive the compliance rules which governs $P$:

$$\Gamma =r_1,r_2,r_3,\dots ,r_n$$

• I.e., The rules for AML compliance could be:

  • Sender & receiver is not in the OFAC list
  • Sender & receiver is not in the FATF list
  • Transaction amount is less than 1000 USD.

• These rules are verified against a set of:

  • AML compliant transactions ($X$)
  • non-AML compliant transactions ($Y$).

A category is a translation of $P(x,\Gamma )$ wherefeatures & thresholds represents the atomic rules $r_i$ of $\Gamma$:

• $\Gamma :$ AML
  • Category: AML_COMPLIANT
• $r_1:$ Sender & receiver is not in the OFAC list
  • Feature: OFAC_LIST_MEMBERSHIP
  • Threshold: false
  • OFAC_LIST_MEMBERSHIP is false for a record.
• $r_1:$ Sender & receiver is not in the FATF list
  • Feature: FATF_LIST_MEMBERSHIP
  • Threshold: false
  • FATF_LIST_MEMBERSHIP is false for a record.
• $r_1:$ Transaction amount is less than 1000 USD.
  • Feature: TRANSACTION_AMOUNT
  • Threshold: 1000
  • TRANSACTION_AMOUNT is less than 1000 for a record.

This translation (see 4.2 Feature Types and Formats) results in a JSON Schema document such as:

{
  "$schema": "http://json-schema.org/draft-07/schema#",
  "$id": "tag:0xbow.io,2024:categories:AML_COMPLIANT",
  "title": "Record is AML Compliant",
  "type": "object",
  "properties": {
    "features": {
      "type": "object",
      "properties": {
        "OFAC_LIST_MEMBERSHIP": {
          "$id": "tag:0xbow.io,2024:categories:AML_COMPLIANT:features:OFAC_LIST_MEMBERSHIP",
          "type": "boolean",
          "default": "true"
        },
        "FATF_LIST_MEMBERSHIP": {
          "$id": "tag:0xbow.io,2024:categories:AML_COMPLIANT:features:FATF_LIST_MEMBERSHIP",
          "type": "boolean",
          "default": "true"
        },
        "TRANSACTION_AMOUNT": {
          "$id": "tag:0xbow.io,2024:categories:AML_COMPLIANT:features:TRANSACTION_AMOUNT",
          "type": "integer",
          "default": "1000"
        }
      },
      "required": [
        "OFAC_LIST_MEMBERSHIP",
        "FATF_LIST_MEMBERSHIP",
        "TRANSACTION_AMOUNT"
      ]
    }
  },
  "required": ["features"]
}

Note that default values are set to invalidate the category for the record.

The default document is a document which satisfy the Schema but where all features have values set to its default value.

{
  "features": {
    "OFAC_LIST_MEMBERSHIP": true,
    "FATF_LIST_MEMBERSHIP": true,
    "TRANSACTION_AMOUNT": 1000
  }
}

As outlined in section 4.3, the Feature Extractor delivers the values of these features in the form of JSON patch operations.

"patch": [
  {
    "op": 2,
    "root": "0x010010",
    "$id": "tag:0xbow.io,2024:categories:AML_COMPLIANT:features:OFAC_LIST_MEMBERSHIP",
    "value": "false",
    "merkle-proof": {}
  },
  {
    "op": 2,
    "root": "0x010010",
    "$d": "tag:0xbow.io,2024:categories:AML_COMPLIANT:features:FATF_LIST_MEMBERSHIP",
    "value": "false",
    "merkle-proof": {}
  },
  {
    "op": 2,
    "root": "0x010010",
    "$d": "tag:0xbow.io,2024:categories:AML_COMPLIANT:features:TRANSACTION_AMOUNT",
    "value": "100",
    "merkle-proof": {}
  }
]

If when applying these patches to the default document, the document still satisfies the schema and the features are consistent with the rules, then the record is classified as AML_COMPLIANT.

{
  "features": {
    "OFAC_LIST_MEMBERSHIP": false,
    "FATF_LIST_MEMBERSHIP": false,
    "TRANSACTION_AMOUNT": 100
  }
}