# Applied Threat Intelligence Fusion Feed

Supported in:

The Applied Threat Intelligence (ATI) Fusion Feed is a collection of Indicators of Compromise
(IoCs), including hashes, IPs, domains, and URLs, that are associated with known
threat actors, malware strains, active campaigns, and finished intelligence reporting. The feed also includes IoCs from open-source feeds that Mandiant Intelligence has carefully checked and validated, which maximizes value and provides high accuracy.

Mandiant's curation process includes the following phases:

- Frontline incident response: While investigating breaches, Mandiant analysts gain first-hand knowledge of
  attacker tools and techniques.

- Threat research: Dedicated teams track threat actors, analyze malware, and
  uncover emerging attack infrastructure.

- Contextualization: IoCs are mapped to specific threats and campaigns, which
  helps in understanding and prioritizing incidents.

The Breach Analytics feed builds upon the ATI Fusion Feed by including indicators from new and active Mandiant breach investigations. It provides real-time insights into the latest attack trends. To enhance indicator matching, YARA-L rules can use contextual information from the ATI Fusion Feed, such as associated threat groups, an indicator's presence in compromised environments, or Mandiant's automated maliciousness score.

## Write YARA-L rules with the ATI Fusion Feed

Writing YARA-L rules that use the ATI Fusion Feed in Google Security Operations follows a similar process to writing YARA-L rules that use other context entity sources. For more information, see [Create context-aware analytics](https://docs.cloud.google.com/chronicle/docs/detection/context-aware-analytics).

> [!NOTE]
> **Note:** If [Data RBAC](https://docs.cloud.google.com/chronicle/docs/administration/datarbac-overview) is enabled in your system, select **bind to scope** to test your rule.

### Events and match sections

To write a rule, follow these steps:
1. Filter the selected context entity graph. In this case, `Fusion Feed`.
1. Filter a specific indicator type. For example, `FILE`. See the following example:

    events:
       $context_graph.graph.metadata.product_name = "MANDIANT_FUSION_IOC"
       $context_graph.graph.metadata.vendor_name = "MANDIANT_FUSION_IOC"
       $context_graph.graph.metadata.source_type = "GLOBAL_CONTEXT"
       $context_graph.graph.metadata.entity_type = "FILE"

You can add any other conditions of the event or context entity in the `events` section. You can join a field from the context entity and UDM event field.

In the following example, the placeholder variable `ioc` is used to do a transitive join between the context entity and the event. The `ioc` variable is then used in the `match` section to make sure there's a match in a specific timeframe.

       $ioc = $context_graph.graph.entity.file.md5
       $ioc = $e1.principal.process.file.md5

    match:
       $ioc over 1h

> [!NOTE]
> **Note:** The system doesn't use the `MANDIANT_FUSION_IOC` log source for automatic IOC matching. This source lets you use these IOCs within your rules.

For more information about context entity fields that can be used in YARA-L rules, see the [Fusion Feed context entity fields](https://docs.cloud.google.com/chronicle/docs/detection/ati-fusion-feed#Fusionfeedcontextentity) section.

### Outcome section

Continuing with the previous example, the basic indicator matching rule is set up
against file hashes in the context entities in the `graph.entity.file.md5`
field and `principal.process.file.md5` UDM field.

Because this rule can match a large number of events, it is recommended that you refine the rule to match on context entities that have specific intelligence. For example, you may want to match on the confidence score that Mandiant assigned to the indicator, whether it was seen in a breached environment or the malware family associated with the indicator. This can all be done in the `outcome` section of the rule.

     outcome:
       // Extract the Mandiant Automated Intel confidence score of maliciousness
       $confidence_score = max(if($context_graph.graph.metadata.threat.verdict_info.source_provider = "Mandiant Automated Intel", $context_graph.graph.metadata.threat.verdict_info.confidence_score, 0))
       // Extract the status of the indicator as seen in a breached environment
       $breached = max(if($context_graph.graph.metadata.threat.verdict_info.pwn = true, 1, 0))

       // Intermediary outcome variable to combine conditions of intelligence extracted in the previous outcome variables.
       // Return 1 if conditions are met, otherwise return 0.
       $matched_conditions = if($confidence_score >= 80 AND $breached = 1, 1, 0)

In the `outcome` section of the YARA-L rule, the confidence score is
extracted using an `if statement` wrapped in a `max` function. This technique is
required for multi-event rules. The same technique is used to extract the
`pwn` variable from `verdict_info`, which indicates whether an
indicator has been seen in a breached environment identified by Mandiant.

These two outcome variables are then combined together in another
`matched_conditions` variable, which allows the use of chained logic
in the `condition` section.

### Condition section

The `condition` section ensures that `e1`,
`context_graph`, and `matched_conditions` exist and or
match the specified condition.

     condition:
       // Ensure $e1, $context_graph and $matched_conditions conditions are met.
       $e1 AND $context_graph AND $matched_conditions = 1

### Complete YARA-L rule

At this point the rule is ready for use and should look like the following:

    rule fusion_feed_example_principal_process_file_md5 {
     meta:
       rule_name = "File Hash - Applied Threat Intelligence"
       description = "Matches file hashes against the Applied Threat Intelligence Fusion Feed."

     events:
       // Filter graph
       $context_graph.graph.metadata.product_name = "MANDIANT_FUSION_IOC"
       $context_graph.graph.metadata.vendor_name = "MANDIANT_FUSION_IOC"
       $context_graph.graph.metadata.entity_type = "FILE"
       $context_graph.graph.metadata.source_type = "GLOBAL_CONTEXT"

       // Do join
       $ioc = $context_graph.graph.entity.file.md5
       $ioc = $e1.principal.process.file.md5

     match:
       $ioc over 1h

     outcome:
       // Extract the Mandiant Automated Intel confidence score of maliciousness
       $confidence_score = max(if($context_graph.graph.metadata.threat.verdict_info.source_provider = "Mandiant Automated Intel", $context_graph.graph.metadata.threat.verdict_info.confidence_score, 0))
       // Extract the status of the indicator as seen in a breached environment
       $breached = max(if($context_graph.graph.metadata.threat.verdict_info.pwn = true, 1, 0))

       // Intermediary outcome variable to combine conditions of intelligence extracted in the previous outcome variables.
       // Return 1 if conditions are met, otherwise return 0.
       $matched_conditions = if($confidence_score >= 80 AND $breached = 1, 1, 0)

     condition:
       // Ensure $e1, $context_graph and $matched_conditions conditions are met.
       $e1 AND $context_graph AND $matched_conditions = 1
    }

### ATI Fusion Feed context entity fields

You can use many fields from the ATI Fusion Feed in rules. These fields
are all defined in the [Unified Data Model field list](https://docs.cloud.google.com/chronicle/docs/reference/udm-field-list).
The following fields are relevant for prioritizing indicators:

| Entity field | Possible values |
|---|---|
| `metadata.threat.associations.type` | `MALWARE`, `THREAT_ACTOR` |
| `metadata.threat.associations.name` | Threat association name |
| `metadata.threat.verdict_info.pwn` | `TRUE`, `FALSE` |
| `metadata.threat.verdict_info.pwn_first_tagged_time.seconds` | Timestamp (seconds) |

Some fields have key-value pairs that need to be used in combination to access
the correct values. For example:

| Entity field 1 | Values | Entity field 2 | Values |
|---|---|---|---|
| `metadata.threat.verdict_info.source_provider` | Mandiant Global Intel | `metadata.threat.verdict_info.global_hits_count` | Integer |
| `metadata.threat.verdict_info.source_provider` | Mandiant Global Intel | `metadata.threat.verdict_info.global_customer_count` | Integer |
| `metadata.threat.verdict_info.source_provider` | Mandiant Analyst Intel | `metadata.threat.verdict_info.confidence_score` | Integer |
| `metadata.threat.verdict_info.source_provider` | Mandiant Automated Intel | `metadata.threat.verdict_info.confidence_score` | Integer |

In the `outcome` section of a YARA-L rule, you can access a value designated
by a specific key using the following command:

    $hit_count = max(if($context_graph.graph.metadata.threat.verdict_info.source_provider = "Mandiant Global Intel", $context_graph.graph.metadata.threat.verdict_info.global_hits_count, 0))

Examining entity matches in Google Security Operations helps you gain a comprehensive view of
the data, revealing additional fields that can be valuable in assessing the
priority and context of an indicator alert.

The following example shows a Fusion Feed context entity as an initial reference point:

    {
      "metadata": {
        "product_entity_id": "md5--147d19e6-cdae-57bb-b9a1-a8676265fa4c",
        "collected_timestamp": {
          "seconds": "1695165683",
          "nanos": 48000000
        },
        "vendor_name": "MANDIANT_FUSION_IOC",
        "product_name": "MANDIANT_FUSION_IOC",
        "product_version": "1710194393",
        "entity_type": "FILE",
        "creation_timestamp": {
          "seconds": "1710201600"
        },
        "interval": {
          "start_time": {
            "seconds": "1"
          },
          "end_time": {
            "seconds": "253402300799"
          }
        },
        "threat": [
          {
            "category_details": [
              "A phishing email message or the relevant headers from a phishing email."
            ],
            "severity_details": "HIGH",
            "confidence_details": "75",
            "risk_score": 75,
            "first_discovered_time": {
              "seconds": "1683294326"
            },
            "associations": [
              {
                "id": "threat-actor--3e5e6bdf-5b4e-5166-84fa-83045e637f23",
                "type": "THREAT_ACTOR",
                "name": "UNC2633"
              },
              {
                "id": "threat-actor--3e5e6bdf-5b4e-5166-84fa-83045e637f23",
                "country_code": [
                  "unknown"
                ],
                "type": "THREAT_ACTOR",
                "name": "UNC2633",
                "description": "UNC2633 is a distribution threat cluster that delivers emails containing malicious attachments or links that lead to malware payloads, primarily QAKBOT, but also SNOWCONE.GZIPLOADER (which leads to ICEDID) and MATANBUCHUS. Historically, UNC2633 has distributed ZIP files containing malicious Excel files that download malware payloads. In early 2023, UNC2633 started distributing OneNote files (.one) that usually led to QAKBOT. It has also leveraged HTML smuggling to distribute ZIP files containing IMG files that contain LNK files and malware payloads.",
                "alias": [
                  {
                    "name": "TA570 (Proofpoint)"
                  }
                ],
                "first_reference_time": {
                  "seconds": "1459085092"
                },
                "last_reference_time": {
                  "seconds": "1687392000"
                },
                "industries_affected": [
                  "Aerospace & Defense",
                  "Agriculture",
                  "Automotive",
                  "Chemicals & Materials",
                  "Civil Society & Non-Profits",
                  "Construction & Engineering",
                  "Education",
                  "Energy & Utilities",
                  "Financial Services",
                  "Governments",
                  "Healthcare",
                  "Hospitality",
                  "Insurance",
                  "Legal & Professional Services",
                  "Manufacturing",
                  "Media & Entertainment",
                  "Oil & Gas",
                  "Pharmaceuticals",
                  "Retail",
                  "Technology",
                  "Telecommunications",
                  "Transportation"
                ]
              }
            ],
            "campaigns": [
              "CAMP.23.007"
            ],
            "last_updated_time": {
              "seconds": "1695165683",
              "nanos": 48000000
            },
            "verdict_info": [
              {
                "source_provider": "Mandiant Automated Intel",
                "confidence_score": 75
              },
              {
                "verdict_type": "ANALYST_VERDICT",
                "confidence_score": 75
              },
              {
                "source_count": 91,
                "response_count": 1,
                "verdict_type": "PROVIDER_ML_VERDICT",
                "malicious_count": 1,
                "ioc_stats": [
                  {
                    "ioc_stats_type": "MANDIANT_SOURCES",
                    "second_level_source": "Knowledge Graph",
                    "quality": "HIGH_CONFIDENCE",
                    "malicious_count": 1,
                    "response_count": 1,
                    "source_count": 8
                  },
                  {
                    "ioc_stats_type": "MANDIANT_SOURCES",
                    "second_level_source": "Malware Analysis",
                    "source_count": 4
                  },
                  {
                    "ioc_stats_type": "MANDIANT_SOURCES",
                    "second_level_source": "Spam Monitoring",
                    "source_count": 1
                  },
                  {
                    "ioc_stats_type": "THIRD_PARTY_SOURCES",
                    "second_level_source": "Crowdsourced Threat Analysis",
                    "source_count": 71
                  },
                  {
                    "ioc_stats_type": "THIRD_PARTY_SOURCES",
                    "first_level_source": "MISP",
                    "second_level_source": "Trusted Software List",
                    "source_count": 3
                  },
                  {
                    "ioc_stats_type": "THIRD_PARTY_SOURCES",
                    "first_level_source": "Threat Intelligence Feeds",
                    "second_level_source": "Digitalside It Hashes",
                    "source_count": 1
                  },
                  {
                    "ioc_stats_type": "THIRD_PARTY_SOURCES",
                    "first_level_source": "Threat Intelligence Feeds",
                    "second_level_source": "Tds Harvester",
                    "source_count": 1
                  },
                  {
                    "ioc_stats_type": "THIRD_PARTY_SOURCES",
                    "first_level_source": "Threat Intelligence Feeds",
                    "second_level_source": "Urlhaus",
                    "source_count": 1
                  }
                ]
              },
              {
                "source_provider": "Mandiant Analyst Intel",
                "confidence_score": 75,
                "pwn": true,
                "pwn_first_tagged_time": {
                  "seconds": "1683911695"
                }
              }
            ],
            "last_discovered_time": {
              "seconds": "1683909854"
            }
          }
        ],
        "source_type": "GLOBAL_CONTEXT",
        "source_labels": [
          {
            "key": "is_scanner",
            "value": "false"
          },
          {
            "key": "osint",
            "value": "false"
          },
          {
            "key": "misp_akamai",
            "value": "false"
          },
    ...
          {
            "key": "has_pwn",
            "value": "2023-05-12T17:14:55.000+0000"
          }
        ],
        "event_metadata": {
          "id": "\\000\\000\\000\\000\\034Z\\n\\2545\\237\\367\\353\\271\\357\\302\\215t\\330\\275\\237\\000\\000\\000\\000\\007\\000\\000\\000\\206\\000\\000\\000",
          "base_labels": {
            "log_types": [
              "MANDIANT_FUSION_IOC"
            ],
            "allow_scoped_access": true
          }
        }
      },
      "entity": {
        "file": {
          "sha256": "000bc5900dc7a32851e380f418cc178ff0910242ee0561ae37ff424e6d3ec64a",
          "md5": "f0095b0a7480c826095d9ffc9d5d2d8f",
          "sha1": "8101315b9fbbf6a72bddbfe64837d246f4c8b419"
        },
        "labels": [
          {
            "key": "is_scanner",
            "value": "false"
          },
          {
            "key": "osint",
            "value": "false"
          },
          {
            "key": "misp_akamai",
            "value": "false"
          },
    ...
        ]
      }
    }

### Complex conditions

To use multiple fields in a context entity, you can combine
multiple outcome variables to create more complex conditional logic. Intermediary outcome variables can be used to combine multiple fields. These variables are then combined to form a new outcome variable that can be used in the `condition` section.

For example:

    // Value will be 1 if threat.associations.type = "MALWARE"
    // Wrapper max function required for multi-event rules
    $is_attributed_malware = max(if($entity_context.graph.metadata.threat.associations.type = "MALWARE", 1, 0))

    // Value will be 1 if threat.associations.type = "THREAT_ACTOR"
    $is_attributed_actor = max(if($entity_context.graph.metadata.threat.associations.type = "THREAT_ACTOR", 1,0))

    // Value will be the sum of the $is_attributed_malware $is_attributed_malware and $is_attributed_actor
    $is_attributed = if($is_attributed_malware = 1, 1, 0)
                        +
                        if($is_attributed_actor = 1, 1, 0)

    // If the value of $is_attributed is greater than 1, this indicates the indicator has been attributed at least once with the type "MALWARE" or "THREAT_ACTOR"

In this example, two intermediary outcome variables, `is_attributed_malware`
and `is_attributed_actor`, are combined in an outcome variable
`is_attributed`.

The intermediary outcome values return numerical values, which
allows numerical comparisons in the new outcome variable.

The value in `is_attributed` will be 1 or greater if
the indicator has at least one threat association of type `MALWARE`
or `THREAT_ACTOR`.

### Flexible joins in YARA-L rule

To reduce the number of rules required, you can use flexible joins between IoCs to connect multiple UDM fields to a context entity.

The following example shows the use of flexible joins for multiple UDM fields in the `event` section:

      events:
        // Filter graph
        $mandiant.graph.metadata.product_name = "MANDIANT_FUSION_IOC"
        $mandiant.graph.metadata.vendor_name = "MANDIANT_FUSION_IOC"
        $mandiant.graph.metadata.entity_type = "FILE"
        $mandiant.graph.metadata.source_type = "GLOBAL_CONTEXT"

        $mandiant.graph.entity.file.md5 = strings.coalesce($e.target.process.file.md5, $e.target.process.file.md5) OR
        $mandiant.graph.entity.file.md5 = strings.coalesce($e.principal.process.file.md5, $e.principal.process.file.md5)

## Migrate Mandiant legacy feeds to GTI

Google SecOps is consolidating threat intelligence data sources. To
provide a broader set of Indicators of Compromise (IoCs) and simplify integration,
we're migrating from legacy Mandiant feeds to the unified
Google Threat Intelligence (GTI) feed.

This guide is for detection engineers and analysts who want to maintain
uninterrupted threat coverage while migrating to the unified GTI feed. It
explains how to update existing YARA-L rules to match the new GTI IoCs while
preserving the legacy alert functionality. By updating your rule
syntax using the provided timeline and migration workflows, you ensure a smooth
transition before legacy feeds are frozen.

> [!NOTE]
> **Note:** The method for how Google SecOps ingests and processes IoCs from the `GTI_IOC` feed differs slightly from the method for legacy feed types. If you notice discrepancies in the number of IOCs or rule matches, it may be due to these differences. For more information, see [Understand GTI_IOC feed ingestion filtering](https://docs.cloud.google.com/chronicle/docs/detection/ati-fusion-feed#understand-gti-ioc-feed-ingestion-filtering).

### Update your YARA-L rules

We recommend that you update your YARA-L rules to the new GTI product name to
maintain uninterrupted coverage.

| **Legacy Mandiant feed name** | **New GTI product name** | **Licensing tier** |
|---|---|---|
| `MANDIANT_FUSION_IOC` | GTI_IOC | Enterprise or Enterprise+ |
| `OPEN_SOURCE_INTEL_IOC` | GTI_IOC | Enterprise or Enterprise+ |
| `MANDIANT_ACTIVE_BREACH_IOC` | GTI_IOC | Enterprise+ |

### YARA-L migration reference

Use these snippets to manually update your rules if you prefer to *not* use the
Gemini **Apply** function.

#### Basic migration

Use this YARA-L syntax to match against both historical data and new GTI data.

**Old YARA-L syntax:**

    $mandiant.graph.metadata.product_name = "MANDIANT_FUSION_IOC"

**Recommended YARA-L syntax:**

    (
        $mandiant.graph.metadata.product_name = "GTI_IOC"
        $mandiant.graph.metadata.threat_intel.stable = true
    )

> [!NOTE]
> **Note:** The `GTI_IOC` feed contains both `stable` and *unreviewed* IoCs. IoCs marked as `stable` are verified and have a high confidence score, whereas unreviewed IoCs are pending verification. For more information, see [Understand GTI_IOC feed ingestion filtering](https://docs.cloud.google.com/chronicle/docs/detection/ati-fusion-feed#understand-gti-ioc-feed-ingestion-filtering).

### Understand the migration workflow

The following workflow reflects the updated Rules Management experience designed
to guide you through the migration.

1. **Identify impacted rules.** Open the **Rules Management/Lister** page. A prominent global banner appears if your tenant has impacted rules.
   - **Indicator:** Rules referencing legacy feeds display a warning icon or an `Update Required` tooltip.
   - **Action:** Click **Ask Gemini to help update** or select a specific rule to see details.
2. **Analyze the rule details** . When you open an impacted rule, the rule editor
   highlights the legacy syntax (for example, `$mandiant.graph.metadata.product_name = "MANDIANT_FUSION_IOC"`).

   A dedicated migration button appears in the editor header, which triggers the
   Gemini side panel.
3. **Read the Gemini-assisted update**. The Gemini side panel automatically
   loads a pre-populated context card:

   - Gemini explains the mapping between your old feed and the new GTI feed based on your entitlement.
   - Gemini provides a **Suggested Fix** code block. Review the updated YARA-L logic, which adds `GTI_IOC` while maintaining the legacy reference for historical coverage.
4. Click **Apply to Editor** to swap the old logic for the new suggested syntax.

5. **Verify your new syntax** . Once the rule is saved with the updated `GTI_IOC`
   syntax, the following details appear:

   - The **Update Required** warning icon disappears from the editor and the main **Rules Lister** page.
   - Your migration status updates in real-time as you clear the flagged rules.

## Understand GTI_IOC feed ingestion filtering

To ensure high fidelity and minimize noise, Google SecOps filters IoCs in the `GTI_IOC` feed. Google SecOps includes an IoC from the feed only if it meets at least one of the following criteria:

- It has a GTI confidence score of 80 or higher.

  > [!NOTE]
  > **Note:** You can retrieve the confidence score (that is, `gti_confidence_score`) using the [GTI API](https://gtidocs.virustotal.com/reference/ip-object).

- It has a GTI verdict of malicious.

- It has a known threat association (for example, it is linked to a specific threat campaign).

Unreviewed IoCs that meet these criteria are still ingested into Google SecOps and are available for use in your YARA-L rules, but they are marked with `metadata.threat.stable = false`. Once these IoCs have been fully reviewed or validated, the `stable` tag is updated to `true`.

For more information, see [Google Threat Intelligence Indicator Score](https://gtidocs.virustotal.com/docs/google-threat-intelligence-indicator-score).

**Need more help?** [Get answers from Community members and Google SecOps professionals.](https://security.googlecloudcommunity.com/)