Kamal Ndousse

19 papers A* 2Journal 15Unranked 2
YearRankTypeTitle / Venue / Authors
2024 J jnl
CoRR
Evan Hubinger, Carson Denison, Jesse Mu, Mike Lambert, Meg Tong, Monte MacDiarmid, Tamera Lanham, Daniel M. Ziegler, Tim Maxwell, Newton Cheng, Adam S. Jermyn, Amanda Askell, Ansh Radhakrishnan, Cem Anil, David Duvenaud, Deep Ganguli, Fazl Barez, Jack Clark, Kamal Ndousse, Kshitij Sachan, Michael Sellitto, Mrinank Sharma, Nova DasSarma, Roger B. Grosse, Shauna Kravec, Yuntao Bai, Zachary Witten, Marina Favaro, Jan Brauner, Holden Karnofsky, Paul F. Christiano, Samuel R. Bowman, Logan Graham, Jared Kaplan, Sören Mindermann, Ryan Greenblatt, Buck Shlegeris, Nicholas Schiefer, Ethan Perez
2024 A* conf
ICLR
Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R. Bowman, Esin Durmus, Zac Hatfield-Dodds, Scott R. Johnston, Shauna Kravec, Timothy Maxwell, Sam McCandlish, Kamal Ndousse, Oliver Rausch, Nicholas Schiefer, Da Yan, Miranda Zhang, Ethan Perez
2023 conf
ACL (Findings)
Ethan Perez, Sam Ringer, Kamile Lukosiute, Karina Nguyen, Edwin Chen, Scott Heiner, Craig Pettit, Catherine Olsson, Sandipan Kundu, Saurav Kadavath, Andy Jones, Anna Chen, Benjamin Mann, Brian Israel, Bryan Seethor, Cameron McKinnon, Christopher Olah, Da Yan, Daniela Amodei, Dario Amodei, Dawn Drain, Dustin Li, Eli Tran-Johnson, Guro Khundadze, Jackson Kernion, James Landis, Jamie Kerr, Jared Mueller, Jeeyoon Hyun, Joshua Landau, Kamal Ndousse, Landon Goldberg, Liane Lovitt, Martin Lucas, Michael Sellitto, Miranda Zhang, Neerav Kingsland, Nelson Elhage, Nicholas Joseph, Noemí Mercado, Nova DasSarma, Oliver Rausch, Robin Larson, Sam McCandlish, Scott Johnston, Shauna Kravec, Sheer El Showk, Tamera Lanham, Timothy Telleen-Lawton, Tom Brown, Tom Henighan, Tristan Hume, Yuntao Bai, Zac Hatfield-Dodds, Jack Clark, Samuel R. Bowman, Amanda Askell, Roger B. Grosse, Danny Hernandez, Deep Ganguli, Evan Hubinger, Nicholas Schiefer, Jared Kaplan
2023 J jnl
CoRR
Sandipan Kundu, Yuntao Bai, Saurav Kadavath, Amanda Askell, Andrew Callahan, Anna Chen, Anna Goldie, Avital Balwit, Azalia Mirhoseini, Brayden McLean, Catherine Olsson, Cassie Evraets, Eli Tran-Johnson, Esin Durmus, Ethan Perez, Jackson Kernion, Jamie Kerr, Kamal Ndousse, Karina Nguyen, Nelson Elhage, Newton Cheng, Nicholas Schiefer, Nova DasSarma, Oliver Rausch, Robin Larson, Shannon Yang, Shauna Kravec, Timothy Telleen-Lawton, Thomas I. Liao, Tom Henighan, Tristan Hume, Zac Hatfield-Dodds, Sören Mindermann, Nicholas Joseph, Sam McCandlish, Jared Kaplan
2023 J jnl
CoRR
Deep Ganguli, Amanda Askell, Nicholas Schiefer, Thomas I. Liao, Kamile Lukosiute, Anna Chen, Anna Goldie, Azalia Mirhoseini, Catherine Olsson, Danny Hernandez, Dawn Drain, Dustin Li, Eli Tran-Johnson, Ethan Perez, Jackson Kernion, Jamie Kerr, Jared Mueller, Joshua Landau, Kamal Ndousse, Karina Nguyen, Liane Lovitt, Michael Sellitto, Nelson Elhage, Noemí Mercado, Nova DasSarma, Oliver Rausch, Robert Lasenby, Robin Larson, Sam Ringer, Sandipan Kundu, Saurav Kadavath, Scott Johnston, Shauna Kravec, Sheer El Showk, Tamera Lanham, Timothy Telleen-Lawton, Tom Henighan, Tristan Hume, Yuntao Bai, Zac Hatfield-Dodds, Ben Mann, Dario Amodei, Nicholas Joseph, Sam McCandlish, Tom Brown, Christopher Olah, Jack Clark, Samuel R. Bowman, Jared Kaplan
2023 J jnl
CoRR
Mrinank Sharma, Meg Tong, Tomasz Korbak, David Duvenaud, Amanda Askell, Samuel R. Bowman, Newton Cheng, Esin Durmus, Zac Hatfield-Dodds, Scott R. Johnston, Shauna Kravec, Timothy Maxwell, Sam McCandlish, Kamal Ndousse, Oliver Rausch, Nicholas Schiefer, Da Yan, Miranda Zhang, Ethan Perez
2022 J jnl
CoRR
Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, Carol Chen, Catherine Olsson, Christopher Olah, Danny Hernandez, Dawn Drain, Deep Ganguli, Dustin Li, Eli Tran-Johnson, Ethan Perez, Jamie Kerr, Jared Mueller, Jeffrey Ladish, Joshua Landau, Kamal Ndousse, Kamile Lukosiute, Liane Lovitt, Michael Sellitto, Nelson Elhage, Nicholas Schiefer, Noemí Mercado, Nova DasSarma, Robert Lasenby, Robin Larson, Sam Ringer, Scott Johnston, Shauna Kravec, Sheer El Showk, Stanislav Fort, Tamera Lanham, Timothy Telleen-Lawton, Tom Conerly, Tom Henighan, Tristan Hume, Samuel R. Bowman, Zac Hatfield-Dodds, Ben Mann, Dario Amodei, Nicholas Joseph, Sam McCandlish, Tom Brown, Jared Kaplan
2022 J jnl
CoRR
Ethan Perez, Sam Ringer, Kamile Lukosiute, Karina Nguyen, Edwin Chen, Scott Heiner, Craig Pettit, Catherine Olsson, Sandipan Kundu, Saurav Kadavath, Andy Jones, Anna Chen, Ben Mann, Brian Israel, Bryan Seethor, Cameron McKinnon, Christopher Olah, Da Yan, Daniela Amodei, Dario Amodei, Dawn Drain, Dustin Li, Eli Tran-Johnson, Guro Khundadze, Jackson Kernion, James Landis, Jamie Kerr, Jared Mueller, Jeeyoon Hyun, Joshua Landau, Kamal Ndousse, Landon Goldberg, Liane Lovitt, Martin Lucas, Michael Sellitto, Miranda Zhang, Neerav Kingsland, Nelson Elhage, Nicholas Joseph, Noemí Mercado, Nova DasSarma, Oliver Rausch, Robin Larson, Sam McCandlish, Scott Johnston, Shauna Kravec, Sheer El Showk, Tamera Lanham, Timothy Telleen-Lawton, Tom Brown, Tom Henighan, Tristan Hume, Yuntao Bai, Zac Hatfield-Dodds, Jack Clark, Samuel R. Bowman, Amanda Askell, Roger B. Grosse, Danny Hernandez, Deep Ganguli, Evan Hubinger, Nicholas Schiefer, Jared Kaplan
2022 J jnl
CoRR
Joel Lehman, Jonathan Gordon, Shawn Jain, Kamal Ndousse, Cathy Yeh, Kenneth O. Stanley
2022 J jnl
CoRR
Catherine Olsson, Nelson Elhage, Neel Nanda, Nicholas Joseph, Nova DasSarma, Tom Henighan, Ben Mann, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, Dawn Drain, Deep Ganguli, Zac Hatfield-Dodds, Danny Hernandez, Scott Johnston, Andy Jones, Jackson Kernion, Liane Lovitt, Kamal Ndousse, Dario Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, Chris Olah
2022 J jnl
CoRR
Saurav Kadavath, Tom Conerly, Amanda Askell, Tom Henighan, Dawn Drain, Ethan Perez, Nicholas Schiefer, Zac Hatfield-Dodds, Nova DasSarma, Eli Tran-Johnson, Scott Johnston, Sheer El Showk, Andy Jones, Nelson Elhage, Tristan Hume, Anna Chen, Yuntao Bai, Sam Bowman, Stanislav Fort, Deep Ganguli, Danny Hernandez, Josh Jacobson, Jackson Kernion, Shauna Kravec, Liane Lovitt, Kamal Ndousse, Catherine Olsson, Sam Ringer, Dario Amodei, Tom Brown, Jack Clark, Nicholas Joseph, Ben Mann, Sam McCandlish, Chris Olah, Jared Kaplan
2022 J jnl
CoRR
Samuel R. Bowman, Jeeyoon Hyun, Ethan Perez, Edwin Chen, Craig Pettit, Scott Heiner, Kamile Lukosiute, Amanda Askell, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, Christopher Olah, Daniela Amodei, Dario Amodei, Dawn Drain, Dustin Li, Eli Tran-Johnson, Jackson Kernion, Jamie Kerr, Jared Mueller, Jeffrey Ladish, Joshua Landau, Kamal Ndousse, Liane Lovitt, Nelson Elhage, Nicholas Schiefer, Nicholas Joseph, Noemí Mercado, Nova DasSarma, Robin Larson, Sam McCandlish, Sandipan Kundu, Scott Johnston, Shauna Kravec, Sheer El Showk, Stanislav Fort, Timothy Telleen-Lawton, Tom Brown, Tom Henighan, Tristan Hume, Yuntao Bai, Zac Hatfield-Dodds, Ben Mann, Jared Kaplan
2022 conf
FAccT
Deep Ganguli, Danny Hernandez, Liane Lovitt, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, Nova DasSarma, Dawn Drain, Nelson Elhage, Sheer El Showk, Stanislav Fort, Zac Hatfield-Dodds, Tom Henighan, Scott Johnston, Andy Jones, Nicholas Joseph, Jackson Kernian, Shauna Kravec, Ben Mann, Neel Nanda, Kamal Ndousse, Catherine Olsson, Daniela Amodei, Tom B. Brown, Jared Kaplan, Sam McCandlish, Christopher Olah, Dario Amodei, Jack Clark
2022 J jnl
CoRR
Deep Ganguli, Danny Hernandez, Liane Lovitt, Nova DasSarma, Tom Henighan, Andy Jones, Nicholas Joseph, Jackson Kernion, Benjamin Mann, Amanda Askell, Yuntao Bai, Anna Chen, Tom Conerly, Dawn Drain, Nelson Elhage, Sheer El Showk, Stanislav Fort, Zac Hatfield-Dodds, Scott Johnston, Shauna Kravec, Neel Nanda, Kamal Ndousse, Catherine Olsson, Daniela Amodei, Dario Amodei, Tom B. Brown, Jared Kaplan, Sam McCandlish, Chris Olah, Jack Clark
2022 J jnl
CoRR
Deep Ganguli, Liane Lovitt, Jackson Kernion, Amanda Askell, Yuntao Bai, Saurav Kadavath, Ben Mann, Ethan Perez, Nicholas Schiefer, Kamal Ndousse, Andy Jones, Sam Bowman, Anna Chen, Tom Conerly, Nova DasSarma, Dawn Drain, Nelson Elhage, Sheer El Showk, Stanislav Fort, Zac Hatfield-Dodds, Tom Henighan, Danny Hernandez, Tristan Hume, Josh Jacobson, Scott Johnston, Shauna Kravec, Catherine Olsson, Sam Ringer, Eli Tran-Johnson, Dario Amodei, Tom Brown, Nicholas Joseph, Sam McCandlish, Chris Olah, Jared Kaplan, Jack Clark
2022 J jnl
CoRR
Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, Nicholas Joseph, Saurav Kadavath, Jackson Kernion, Tom Conerly, Sheer El Showk, Nelson Elhage, Zac Hatfield-Dodds, Danny Hernandez, Tristan Hume, Scott Johnston, Shauna Kravec, Liane Lovitt, Neel Nanda, Catherine Olsson, Dario Amodei, Tom B. Brown, Jack Clark, Sam McCandlish, Chris Olah, Benjamin Mann, Jared Kaplan
2021 J jnl
CoRR
Amanda Askell, Yuntao Bai, Anna Chen, Dawn Drain, Deep Ganguli, Tom Henighan, Andy Jones, Nicholas Joseph, Benjamin Mann, Nova DasSarma, Nelson Elhage, Zac Hatfield-Dodds, Danny Hernandez, Jackson Kernion, Kamal Ndousse, Catherine Olsson, Dario Amodei, Tom B. Brown, Jack Clark, Sam McCandlish, Chris Olah, Jared Kaplan
2021 A* conf
ICML
Kamal Ndousse, Douglas Eck, Sergey Levine, Natasha Jaques
2020 J jnl
CoRR
Kamal Ndousse, Douglas Eck, Sergey Levine, Natasha Jaques
APK_FEATURES_PDD.md
← Index APK_FEATURES_PDD.md markdown
# APK Extractor — Product Design Document

**Author:** Engineering Team
**Date:** 2026-02-21
**Status:** Draft
**Target:** redb ingestor pipeline

---

## 1. Overview

This document describes the design for adding APK (Android Package) static analysis to the redb ingestor pipeline. The APK extractor will follow the same architecture used by the existing PE, ELF, and Mach-O extractors: a format-specific base class (`APKExtractor`) with specialized sub-extractors for each analysis dimension.

### 1.1 Goals

- Extract comprehensive static metadata from Android APK files, comparable in depth to our PE/ELF/Mach-O analysis
- Follow the established extractor architecture (base class, sub-extractors, dataclasses, dual-export to Elasticsearch and ClickHouse)
- Enable clustering, hunting, and pivoting on APK-specific fields (permissions, certificates, DEX API usage, package names)
- Integrate with the existing format-agnostic extractors (BasicProperties, Hashes, DIE, CAPA, YARA, Strings, IOC)

### 1.2 Non-Goals

- Dynamic analysis / sandbox execution (out of scope)
- Full DEX decompilation to Java/smali (out of scope; may be a future extension)
- Deep analysis of embedded native `.so` libraries (inventory only; full ELF pipeline deferred)
- Recursive ingestion of APKs embedded inside other APKs (flag only; full recursive ingestion deferred)

---

## 2. Background

### 2.1 What Is an APK

An APK is a ZIP archive containing an Android application. Its internal structure:

| Path | Contents |
|------|----------|
| `AndroidManifest.xml` | Binary Android XML — package name, permissions, components, SDK versions, intent filters |
| `classes.dex` (+ `classes2.dex`, ...) | Dalvik bytecode — compiled Java/Kotlin code |
| `resources.arsc` | Compiled resource table (strings, dimensions, styles) |
| `res/` | Layouts, drawables, raw resources |
| `lib/<abi>/` | Native shared libraries (`.so`) per CPU architecture |
| `META-INF/` | JAR signing (v1 scheme), CERT.RSA/DSA certificates |
| `assets/` | Arbitrary files bundled by the developer |

### 2.2 Current State

The ingestor already detects APK files via Magika (`ingestor.py:1668`), but the handler is a no-op — it logs `"APK file detected"` and returns without running any extractors. All infrastructure for registration, dispatch, and export is already in place.

### 2.3 Industry Reference

This design was informed by analysis of existing platforms:

- **VirusTotal** — Extracts: hashes (MD5, SHA-1, SHA-256, SSDEEP, TLSH, Permhash), Android metadata (package name, SDK versions, main activity), certificate attributes, permissions with danger flags, full component lists (activities, services, receivers, providers), intent filters, and capability indicators ("performs reflection calls", "makes use of telephony related APIs")
- **Koodous** (https://docs.koodous.com/) — Powered by Androguard for static analysis. Extracts: package name, app name, activities, services, receivers, providers, permissions, intent filters, certificate (SHA-1, issuer, subject), hardcoded URLs, SDK versions. Supports YARA rules with an Androguard module for matching on all these fields
- **APKiD** (https://github.com/rednaga/APKiD) — "PEiD for Android". Signature-based identification of compilers, packers, obfuscators, protectors, anti-VM/debug/root techniques. Uses YARA rules on DEX/APK/ELF. Available under GPL or commercial perpetual license (https://github.com/rednaga/APKiD/blob/master/LICENSE.COMMERCIAL)
- **APKDetect** (https://www.apkdetect.com/) — Malware family identification (30+ families), configuration extraction, loader recognition, shared code detection

---

## 3. Architecture

### 3.1 Existing Extractor Pattern

All extractors in redb follow this hierarchy:

```
Extractor (abstract base — redb/extractors/extractor.py)
├── PEExtractor (redb/extractors/pe_extractor.py)
│   ├── PEFeaturesExtractor
│   ├── PEImportExtractor
│   ├── PESectionExtractor
│   └── ...
├── ELFExtractor (redb/extractors/elf_extractor.py)
│   ├── ELFFeaturesExtractor
│   ├── ELFImportExtractor
│   └── ...
├── MachOExtractor (redb/extractors/macho_extractor.py)
│   ├── MachOFeaturesExtractor
│   ├── MachOImportExtractor
│   └── ...
└── [Format-agnostic extractors]
    ├── BasicPropertiesExtractor
    ├── HashExtractor
    ├── DIEExtractor
    ├── CAPAExtractor
    ├── YaraExtractor
    └── StringsExtractor
```

Each concrete extractor implements:
- `tag()` — returns a `Tag` enum value identifying the extraction category
- `extract()` — performs the analysis, returns a dataclass instance or `None`
- `prepare_export_data(exporter_type)` — formats data for Elasticsearch or ClickHouse
- `get_clickhouse_table()` — returns the target ClickHouse table name

The format-specific base class (e.g., `PEExtractor`) handles:
- Accepting a pre-parsed object (e.g., `pe=`) to avoid redundant parsing
- Providing shared helper methods used by multiple sub-extractors
- Passing `precomputed_hashes` to the parent `Extractor.__init__()` when available

The ingestor dispatches extractors by filetype detected via Magika, running format-agnostic extractors (BasicProperties, Hashes, DIE, CAPA, YARA) followed by the format-specific list.

### 3.2 APK Extractor Architecture

```
Extractor
└── APKExtractor (NEW — redb/extractors/apk_extractor.py)
    ├── APKFeaturesExtractor      (redb/extractors/apk_extractors/apk_features.py)
    ├── APKManifestExtractor      (redb/extractors/apk_extractors/apk_manifest.py)
    ├── APKPermissionsExtractor   (redb/extractors/apk_extractors/apk_permissions.py)
    ├── APKSignatureExtractor     (redb/extractors/apk_extractors/apk_signature.py)
    ├── APKDexExtractor           (redb/extractors/apk_extractors/apk_dex.py)
    ├── APKResourceExtractor      (redb/extractors/apk_extractors/apk_resources.py)
    ├── APKNativeLibExtractor     (redb/extractors/apk_extractors/apk_native_libs.py)
    └── APKInconsistencyTestsExtractor (redb/extractors/apk_extractors/apk_inconsistency_tests.py)
```

The `APKExtractor` base class will:
- Accept an optional pre-parsed `androguard.core.apk.APK` object (`apk=`) to share across sub-extractors
- Parse the APK once in `__init__` if not provided
- Provide shared helpers: `_get_manifest()`, `_get_certificates()`, `_list_files()`, `_is_valid_apk()`

### 3.3 Ingestor Integration

In `ingestor.py:process_binary_file()`, the `elif filetype == "apk"` branch will be expanded to:

1. Parse the APK once using Androguard (`APK(filepath)`)
2. Run format-agnostic extractors (BasicProperties, Hashes, DIE, YARA) — same as PE/ELF/Mach-O
3. Run APK-specific extractors with the shared `apk` object
4. Run Strings + IOC extractors if applicable

---

## 4. Extractor Specifications

### 4.1 APKFeaturesExtractor

**Purpose:** Core APK metadata — the equivalent of `PEFeaturesExtractor` or `ELFFeaturesExtractor`.

**Extracted fields:**
- `package_name` — Android package identifier (e.g., `com.example.app`)
- `app_name` — Human-readable application name
- `version_code` — Internal integer version
- `version_name` — Display version string (e.g., `"1.2.3"`)
- `min_sdk_version` — Minimum Android API level
- `target_sdk_version` — Target Android API level
- `compile_sdk_version` — Compile SDK (if available)
- `main_activity` — Launcher activity class name
- `is_debuggable` — Whether `android:debuggable="true"`
- `allow_backup` — Whether `android:allowBackup="true"`
- `uses_cleartext_traffic` — Whether `android:usesCleartextTraffic="true"`
- `supported_abis` — List of ABIs from `lib/` directory (e.g., `["arm64-v8a", "armeabi-v7a"]`)
- `dex_count` — Number of DEX files
- `total_dex_size` — Combined DEX file size in bytes
- `total_file_count` — Total number of files in the APK archive
- `has_native_code` — Whether `lib/` contains `.so` files
- `has_assets` — Whether `assets/` directory is non-empty
- `uses_libraries` — Declared `<uses-library>` entries
- `earliest_content_modification` — Earliest timestamp from ZIP entry metadata
- `latest_content_modification` — Latest timestamp from ZIP entry metadata
- `contains_embedded_apk` — Whether the archive contains nested APK files (flag only)

**Tag:** `APK_FEATURES`
**ClickHouse table:** `redb_apk_features`

### 4.2 APKManifestExtractor

**Purpose:** Full AndroidManifest.xml component enumeration, mirroring what VirusTotal and Koodous display.

**Extracted fields:**
- `activities` — List of Activity class names with `exported` flag
- `services` — List of Service class names with `exported` flag
- `receivers` — List of BroadcastReceiver class names with `exported` flag
- `providers` — List of ContentProvider class names with `exported` flag
- `intent_filters_by_action` — Aggregated list of all intent filter actions
- `intent_filters_by_category` — Aggregated list of all intent filter categories
- `uses_features` — Declared hardware/software features (e.g., `android.hardware.camera`)
- `meta_data` — Key-value pairs from `<meta-data>` elements
- `manifest_xml` — Full decompiled AndroidManifest.xml as plain text

**Tag:** `APK_MANIFEST`
**ClickHouse table:** `redb_apk_manifest` (one row per APK for aggregated data), `redb_apk_components` (one row per component)

### 4.3 APKPermissionsExtractor

**Purpose:** Dedicated permission analysis with protection-level classification and permhash computation.

**Extracted fields:**
- `permissions` — List of requested permissions with protection level (`normal`, `dangerous`, `signature`, `signatureOrSystem`)
- `dangerous_permissions` — Filtered list of dangerous permissions only
- `dangerous_permission_count` — Count of dangerous permissions
- `custom_permissions` — Permissions defined by the app itself (`<permission>` declarations)
- `total_permission_count` — Total number of requested permissions
- `permhash` — SHA-256 of sorted permission list (Mandiant/Google permhash — https://github.com/google/permhash)

The `permhash` value will also be added to the `Hashes` dataclass in `redb/models/dataclasses.py` so it appears alongside `imphash`, `symhash`, etc. in the unified hash record.

**Tag:** `APK_PERMISSIONS`
**ClickHouse table:** `redb_apk_permissions` (one row per permission per APK)

### 4.4 APKSignatureExtractor

**Purpose:** Certificate and signing scheme analysis, analogous to `PESignatureExtractor` and `MachOSignatureExtractor`.

**Extracted fields:**
- `signature_scheme_versions` — Which signing schemes are present (v1 JAR, v2, v3, v4)
- `is_signed` — Whether the APK has a valid signature
- `number_of_certificates` — Certificate count in the chain
- `x509_certificates` — List of certificate details:
  - `subject` (Distinguished Name)
  - `issuer` (Distinguished Name)
  - `serial_number`
  - `valid_from` / `valid_to`
  - `thumbprint_sha1`
  - `thumbprint_sha256`
  - `algorithm`
  - `key_size`
  - `is_self_signed`
- `signer_subject` — Primary signer's subject DN (convenience field)
- `signer_issuer` — Primary signer's issuer DN

**Tag:** `APK_SIGNATURE`
**ClickHouse table:** `redb_apk_signature`

### 4.5 APKDexExtractor

**Purpose:** DEX file analysis — summarized with categorized API usage (not full method enumeration).

**Output structure (per DEX file):**

- `filename` — DEX filename (e.g., `classes.dex`)
- `sha256` — SHA-256 of the DEX file
- `class_count` — Total number of classes
- `method_count` — Total number of methods
- `string_count` — Total number of string constants
- `top_packages` — List of `{package, class_count}` for top-level Java packages
- `api_usage` — Categorized sensitive API calls:
  - `reflection` — `java.lang.reflect.*`, `Class.forName`, etc.
  - `crypto` — `javax.crypto.*`, `java.security.*`
  - `dynamic_loading` — `DexClassLoader`, `PathClassLoader`, `InMemoryDexClassLoader`
  - `telephony` — `TelephonyManager` methods
  - `sms` — `SmsManager`, `SmsReceiver`
  - `network` — `HttpURLConnection`, `OkHttp`, `Volley`, `Retrofit`
  - `native` — `System.loadLibrary`, `Runtime.exec`
  - `device_info` — `Build.*`, `Settings.Secure.ANDROID_ID`, IMEI/IMSI access
  - `file_io` — `FileOutputStream`, `SharedPreferences`, `SQLiteDatabase`
  - `ipc` — `ContentResolver`, `BroadcastReceiver`, `Binder`
- `obfuscation_indicators`:
  - `short_class_names_pct` — Percentage of classes with names <= 2 chars
  - `short_method_names_pct` — Percentage of methods with names <= 2 chars
  - `non_ascii_identifiers` — Count of identifiers with non-ASCII characters
  - `avg_class_name_length` — Average class name length

**Tag:** `APK_DEX`
**ClickHouse table:** `redb_apk_dex` (one row per DEX file), `redb_apk_dex_api_usage` (one row per API category per DEX)

### 4.6 APKResourceExtractor

**Purpose:** Inventory of embedded resources with filetype detection for suspicious content.

**Extracted fields:**
- `total_resource_count` — Total files in `res/` and `assets/`
- `total_resource_size` — Combined size
- `resource_inventory` — List of `{path, size, sha256, filetype_magika}` for each file
- `suspicious_files` — Files detected as ELF, PE, DEX, APK, ZIP, script, or other executable types
- `suspicious_file_count` — Count of suspicious files

**Tag:** `APK_RESOURCES`
**ClickHouse table:** `redb_apk_resources`

### 4.7 APKNativeLibExtractor

**Purpose:** Inventory of native `.so` libraries per ABI. No deep ELF analysis (inventory only).

**Extracted fields:**
- `native_lib_count` — Total `.so` file count
- `abis` — List of ABI directories present (e.g., `["arm64-v8a", "x86"]`)
- `native_libs` — List of `{abi, filename, size, sha256}` per `.so` file
- `known_packer_libs` — Any `.so` files matching known packer/protector library names (e.g., `libjiagu.so`, `libsecexe.so`, `libDexHelper.so`, `libprotectClass.so`)

**Tag:** `APK_NATIVE_LIBS`
**ClickHouse table:** `redb_apk_native_libs`

### 4.8 APKInconsistencyTestsExtractor

**Purpose:** Anomaly and anti-analysis detection, analogous to `PEInconsistencyTestsExtractor`.

**Extracted fields:**
- `test_zip_bomb` — Compression ratio exceeds threshold
- `test_zip_duplicate_entries` — ZIP contains duplicate filenames
- `test_zip_path_traversal` — ZIP entries with `../` path traversal
- `test_zip_suspicious_timestamps` — Timestamps in the future or at epoch (1980-01-01)
- `test_hidden_dex_files` — DEX files outside standard `classes*.dex` naming or in unexpected locations
- `test_manifest_component_mismatch` — Declared components that don't exist in DEX, or undeclared code
- `test_debuggable_release` — `android:debuggable="true"` combined with a release signature
- `test_emulator_detection_strings` — Presence of Build.FINGERPRINT/MANUFACTURER emulator check strings in DEX
- `test_debugger_detection` — Presence of `Debug.isDebuggerConnected()` calls
- `test_root_detection` — Presence of root detection patterns (su binary checks, Superuser.apk)

**Tag:** `APK_INCONSISTENCY_TESTS`
**ClickHouse table:** `redb_apk_inconsistency_tests`

---

## 5. New Dependencies

### 5.1 Required

| Library | Version | License | Purpose |
|---------|---------|---------|---------|
| **androguard** | >=4.1 | Apache 2.0 | Core APK parsing: binary XML manifest, DEX analysis, certificate extraction, permissions |
| **permhash** | latest | Apache 2.0 | Compute permhash (SHA-256 of sorted permissions) for APK clustering |

### 5.2 Already Available (no changes)

| Library | Current Version | Usage |
|---------|----------------|-------|
| `zipfile` (stdlib) | — | APK archive traversal, ZIP anomaly detection |
| `pyelftools` | 0.32 | Could be used for native .so analysis if scope expands |
| `lief` | 0.17.3 | Has `lief.DEX` module; complement to androguard for DEX class/method enumeration |
| `cryptography` | 43.0.3 | Certificate chain validation (used by PE/Mach-O signature extractors) |
| `Pillow` | 10.4.0 | Icon extraction if needed |
| `magika` | 1.0.1 | Filetype detection for embedded resources |

### 5.3 Future Consideration

| Library | License | Purpose | Notes |
|---------|---------|---------|-------|
| **APKiD** | GPL-3.0 or Commercial (perpetual, royalty-free) | Packer/protector/obfuscator identification | "PEiD for Android". Detects compilers (dx, dexlib, r8), packers (AppGuard, DingXiang, JiaguK), obfuscators (DexGuard, BlackObfuscator), protectors (DexProtector, DxShield), anti-VM/debug/root. Commercial license is perpetual and allows binary redistribution: https://github.com/rednaga/APKiD/blob/master/LICENSE.COMMERCIAL |
| **apksigtool** | MIT | APK Signature Scheme v2/v3/v4 verification | androguard handles v1 well; apksigtool provides more thorough v2+ support |
| **quark-engine** | GPL-3.0 | Behavioral analysis scoring | Similar to CAPA but for Android. Stretch goal |

---

## 6. Data Model Changes

### 6.1 New Dataclasses

Add to `redb/models/dataclasses.py`:
- `APKFeatures`
- `APKManifestComponent`
- `APKPermission`
- `APKCertificate`
- `APKCodeSigningInfo`
- `APKDexFile`
- `APKDexApiUsage`
- `APKResource`
- `APKNativeLib`
- `APKInconsistencyTests`

See Tech Annex for full field definitions.

### 6.2 Existing Dataclass Changes

**`Hashes` dataclass** — add:
```python
permhash: Optional[str] = None  # APK: SHA-256 of sorted permissions (Mandiant/Google)
```

### 6.3 Tag Enum Additions

Add to `redb/extractors/enum.py`:
```python
# APK
APK_FEATURES = "apk_features"
APK_MANIFEST = "apk_manifest"
APK_PERMISSIONS = "apk_permissions"
APK_SIGNATURE = "apk_signature"
APK_DEX = "apk_dex"
APK_RESOURCES = "apk_resources"
APK_NATIVE_LIBS = "apk_native_libs"
APK_INCONSISTENCY_TESTS = "apk_inconsistency_tests"
```

---

## 7. Implementation Phases

### Phase 1 — Core Extractors

1. `APKExtractor` base class + `APKFeaturesExtractor`
2. `APKManifestExtractor`
3. `APKPermissionsExtractor` (including permhash)
4. `APKSignatureExtractor`
5. `APKDexExtractor`
6. `APKResourceExtractor`
7. `APKNativeLibExtractor`
8. Ingestor integration (wire up dispatch in `process_binary_file()`)
9. Hashes dataclass update (add `permhash`)

### Phase 2 — Detection and Anomalies

10. `APKInconsistencyTestsExtractor`
11. Wire up existing format-agnostic extractors for APK (YARA, Strings, IOC)

### Phase 3 — Future (out of scope for this PDD)

- APKiD integration for packer/protector detection
- Deep native library analysis (run ELF pipeline on extracted `.so` files)
- Recursive APK ingestion
- androguard-yara module integration for YARA rules matching on APK metadata

---

## 8. Testing Strategy

- Unit tests per extractor using known APK samples (benign + malicious)
- Validate output against VirusTotal reports for the same samples (cross-reference fields)
- Edge cases: split APKs, obfuscated APKs, packed APKs, APKs with no native code, APKs with no DEX, corrupted APKs
- Integration test: full pipeline run (ingest APK, verify all extractors produce output, verify ClickHouse/ES export)