Xygeni Guardrails

Guardrails Specification

This is a guardrail specification for Xygeni scanner. Guardrails can be executed at server side (Audit Analysis), for more information read CI/CD Audit Analysis

When specific rules or special exit codes are required by pipelines or security policies, complex conditions can be defined using guardrail expressions.

Guardrails enables users to customize and specify how a Xygeni command should behave in the presence of certain conditions or criteria, thereby allowing for flexible and tailored handling of failure scenarios.

Guardrail specification is a subset of Xygeni™ XyFlow Language, tailored for guardrail conditions.

Guardrail Convention

A guardrail expression consists of a name or string, followed by a pipe that contains actions limited to the set selected.

A pipe processes a stream of events matching a condition then performing actions on each event individually or on the full set of matching events.

guardrail
    on
      entries_selector
    when
      expressions...
    then
      then-actions...
    else
      else-actions...
    also
      when ... then ... else
  ...
]

ON Clause

The on clause chooses which items from scan report to process. The following are the recognized names with the report matched and the type of the items to process:

type(s)	report	items
inventory, asset, assets	InventoryReport	Asset
deps, scan-deps, dependencies, sca	ScanReport	Dependency
suspectdeps	SuspectDepsReport	SuspectDependency
malware, evidence	MalwareEvidencesReport	MalwareEvidence
secrets, secret	SecretsReport	PotentialSecret
iac, iac_flaws	IacFlawsReport	IacFlaw
cicd, misconf, misconfiguration(s)	MiscReport	Misconfiguration
codetamper, critical_file_modification	CodeTamperReport	CriticalFileModification
compliance	ComplianceReport	Compliance
entry, entries, item, items	Collection, Array, Map	Object or Map.Entry
any	<Any of the above>	<Corresponding type>

Closed Issues are removed from the item list before processing by the guardrails. It require an available connection to xygeni server to retrieve the list of closed issues for the current project if exits at server side. It applies to secrets, misconfigurations, suspectdeps, malware and iac issue types.

Examples:

# guardrail will apply when running a SuspectDepsAnalysis
'guardrail <name> on suspectdeps when <condition> then ...'

# guardrail will apply when running a SecretsAnalysis or MisconfigurationsAnalysis
'guardrail <name> on secrets,misconf when <condition> then ...'

# guardrail will apply on any analysis
'guardrail <name> on any when <condition> then ...'

Item Properties Reference

Each item type available in the on clause exposes a set of properties that can be used in when expressions, function arguments, or actions. These properties are accessible using dot notation ($property.subproperty) and are context-specific.

Default Properties (Common to Most Items)

These properties are available in most issue-related items (e.g., secrets, misconfigurations, iac flaws, malware, suspect dependencies):

• severity: Issue severity (e.g., info, low, high, critical)

• confidence: Detection confidence (e.g., low, medium, high, highest)

• issueKind: Type or classification of the issue (e.g., secret, iac_flaw)

• detector: Name or identifier of the rule or scanner that found the issue

• location.filepath: File path where the issue was detected

• branch: Name of the branch that triggered the scan

• explanation: Description or rationale behind the detection

CodeVulnerability (on: sast)

• language: (e.g., 'java', 'python')

• cve: CVE identifier (e.g., 'CVE-2022-1234')

SuspectDependency (on: suspectdeps)

• dependency.language: Dependency language

• group: Dependency group (e.g., 'com.example')

• name: Dependency name

• version: Version number

MalwareEvidence (on: malware, evidence)

• kind: Kind of evidence (e.g., 'execution', 'file', 'network', 'registry', '_package', 'sensitive_data', 'system')

PotentialSecret (on: secrets)

• tags.verified: Returns true if the secret is verified

• tags.inactive: Returns true if the secret is inactive

IacFlaw (on: iac, iac_flaws)

• framework: IaC framework (e.g., terraform, ansible)

• type: Type of the IaC issue (e.g., appsec, backup_recovery, convention, encryption, gensec, iam, logging, network, secrets)

Misconfiguration (on: cicd, misconf)

• type: Type of the misconfiguration (e.g., cicd_security, network, credentials)

CriticalFileModification (on: codetamper)

• type: Type of the critical file modification (e.g., codeowners, configurations, shell_script, cicd_workflow)

WHEN Clause

The when clause defines the condition for each item to be processed. A list of matched items will be passed to the then clause. In case an else clause is provided, a list of unmatched items will be passed to the else clause.

See Expressions for more details.

THEN Clause

The then clause defines the actions to be executed on the matched items.

If some items are found (issues or evidence) but none is matched neither else clause is provided, an exitcode will be set to 127 to force a non zero exit.

ELSE Clause

The else clause defines the actions to be executed on the unmatched items.

ALSO Clause

The also clause allows to define a nested when-then-else clause that will be evaluated over the matched items.

General Syntax

Keywords

• guardrail: Indicates the beginning of a guardrail definition.

• on: Specifies the triggering events.

• when: Defines the condition for event/item processing.

• then: Specifies the actions to be performed.

• also: Optional keyword for additional conditions and actions.

Identifiers

Represent variable names or properties and follow the standard naming conventions similar to Java identifiers. They can start with a letter or underscore (_), followed by any combination of letters, digits, and underscores.

Delimiters

• []: Square brackets used for Array elements

• [:]: Square brackets with a colon used for Maps elements

• (): Parentheses used for function calls

• ,: Comma used to separate arguments

• :: Colon used on function calls to identify named arguments

Literals

Include integers, floating-point numbers, strings (enclosed in double quotes "…​"), and booleans (true, false).

Operators

Arithmetic operators:

• +, -, *, /, %

• ++, --

Comparison operators:

• <, >

• ⇐, >=

• =, !=, <>

Logical operators:

• and, &&

• or, ||

• not, !

String operators

• starts-with, ends-with, matches, contains

Collection operators

• in, contains

Object operators

• instanceOf, exists

Examples of operators expressions

severity > 'low' (more than)
location.filepath = 'data/config.yml' (equals)
location.filepath starts-with 'data/'
location.filepath matches '.*\.yml$' (regex)
tags contains 'verified'
confidence in ['high', 'highest']

Comments

Comments can be included to provide clarity and documentation within guardrail definitions. Only Single-line comments are provided, it starts with #.

Functions

Functions are invoked using the @ symbol followed by the function name and a list of arguments enclosed in parentheses, e.g., @functionName(arg1, arg2).

Dot Notation

Dot notation can be used to access object properties and allows for specifying conditions or actions based on properties of objects within the guardrail execution context.

'... when report.baseline = true then ...'

Interpolation and Expression Embedding

Variables can be referenced using the $ symbol followed by the variable name, e.g., $report.baseline. $ alone references the current item, e.g., @print($).

$Variable can be used to interpolate values in a string, e.g., "Report is baseline: $report.baseline".

${} can be used to interpolate an expression in a string, e.g., "Total number of secrets: ${@count(report.secrets)}".

Context Objects

Some objects are saved to global context and can be accessed implicitly from expressions:

• report a reference to the current analysis report generated.

• metadata a direct reference to report.metadata properties if present

• context the context itself is saved to global context

Examples:

'... when metadata.baseline = true then ...'

Expressions

Expressions can include variables, literals, functions, and operators combined to evaluate to a Truth in XyFlow. An expression is Truth when y resolved as:

• a true boolean

• a non-zero number

• a non-blank string

• a non-empty array, map or collection

• a non-null object

Actions

Actions are evaluated over each matched item (default) or over the complete list of matched items. To evaluate over the complete list use a kind name with the _all suffix. To select a specific action, use the @ symbol followed by the name of the action.

Example:

on secrets
  when
    severity = 'critical'
  then
    @exitcode(167)'
  else
    print_all:@print("non-critical:" + @count($))'

Built-in Functions

• Aggregate Functions Available aggregate functions like count, sum, avg, min, max, and sort, which can operate on collections or arrays. These functions are crucial for data manipulation within guardrails.

  • @count(collection): Returns the number of elements in a collection, map, array, or the length of a string.

  • @sum(collection): Calculates the sum of numeric values within a collection or array.

  • @avg(collection): Computes the average of numeric values in a collection or array.

  • @min(collection): Finds the minimum numeric value in a collection or array.

  • @max(collection): Identifies the maximum numeric value in a collection or array.

  • @sort(collection): Returns a new collection with elements sorted in their natural order.

• Exitcode Allows the specification of exit codes for guardrails, which can be used to indicate success or failure states in a customizable manner.

  • @exitcode(code): Sets the exit code of the guardrail, enabling custom exit statuses to communicate outcomes to the CI/CD pipeline.

  • @fail(): Sets exit code of the guardrail to value 1, to force an exit with error. It is a simplified version of the @exitcode(1) function.

• Item Functions

  • @is_new(): Returns true if the element is an Issue (Secret, Misconfiguration, IacFlaw, SuspectDependency, or MalwareEvidence) that wasn't present in the baseline analysis of the current project. Also returns true when a newly added Dependency contains Vulnerabilities not found in the baseline analysis.

• Print/Debug Functions Enables printing messages to different outputs (stdout, stderr, or a file), which can be used for logging or debugging within guardrails.

  • @print(expression): Allows printing of specified messages to stdout, stderr, or a file, useful for logging and debugging purposes. Parameters:

    • item: The text to print

    • output: stdout, stderr, or a file path. Defaults to stdout.

  • @dump(msg): Dumps the message to stdout, typically for debugging. Used during dry-run testing of guardrails.

• Scm Functions Functions for interacting with Source Control Management (SCM) systems, currently GitHub, through functions like github.status_check and github.commit_comment. These functions are used for reporting the results of guardrail checks back to SCM systems.

  • @github.status_check(…​): Creates a GitHub commit status, useful for reporting the outcome of a guardrail evaluation directly within a GitHub repository. Parameters:

    • repo: name of the repo (owner/name)

    • sha: commit sha (default last commit)

    • state: error, failure(default), pending, success

    • description: A short description of the status

  • @github.commit_comment(…​): Posts a comment on a specific commit in GitHub, allowing for inline feedback on guardrail checks. Parameters:

    • repo: name of the repo (owner/name

    • sha: commit sha (default last commit)

    • state: error, failure(default), pending, success

    • body: The body of the comment

    • path: Optional path of the file to comment on

    • Position: Optional position in the file to comment on

Guardrails Examples

Exit on Critical Issues

• Condition: Break the build on any issue with a 'critical' severity level.

• Action: Exit with a specific exit code only when 'critical' severity level is found.

   on
     any
   when
     severity = critical
   then
     @exitcode(167)

Exit on Critical Files

• Condition: Break the build on any issue on critical files.

• Action: Exit with a specific exit code only when issue found on critical files.

   on
     any
   when
     severity in ['critical', 'high'] and
     location.filepath starts-with 'src/admin'
   then
     @exitcode(167)

New critical Secret or Dependency Issue not allowed

• Condition: Break the build on any new critical secret or dependency.

• Action: Exit with a specific exit code only when new critical issues not present in baseline analysis are found.

• NOTE: This require a connection to xygeni server in order to retrieve current baseline issues.

   guardrail
     new_critical_issues
   on
     secrets,suspectdeps
   when
     @is_new() and 
     severity = 'critical'
   then
     @exitcode(155)

Critical Secrets Exposure

• Condition: Any issue where the kind is a 'secret', with a 'critical' severity level.

• Action: Break the build with a specific exit code and set commit status failure

   on
     secrets
   when
     severity = critical
   then
     print_status_all:@print("updating commit status on repo: " + metadata.scm.fullName),
     set_status_all:@github.status_check(repo: $metadata.scm.fullName,state: 'failure',
    description: "Secret found in ${report.metadata.projectName}",context: 'security/secret'),
     @exitcode(167)

High Confidence Misconfigurations

• Condition: Issues categorized under 'misconfiguration' with 'high' or 'critical' confidence and 'high' or 'highest' .

• Condition: Issues categorized under 'misconfiguration' with 'high' or 'critical' severity and 'high' or 'highest' confidence.

• Action: Break the build with a specific exit code and open a ticket for the relevant team

# High Confidence Misconfigurations
on
  cicd
when
  severity >= high and confidence >= high
then
  # Creation of a ticket for the relevant team with a request for urgent remediation
  print_action:@print("open ticket with: " + metadata.scm.fullName),
  open_ticket:@http.send(
    url: "http://issueserver/rest/api/issue",
    method: 'POST',
    headers: [tokenId: "AHGFCM..."],
    body: "CI/CD issue found in ${report.metadata.projectName}, file ${location.file}, commitSha: ${commitSha}",
    context: 'security/secret'),
  @exitcode(177)

This guardrail must be run server-side, as a CI/CD Audit.