README

trueseeing is a fast, accurate and resillient vulnerability scanner for iOS/Android apps. We operate on the Dalvik VM level for Android -- i.e. we don't care if the target app is obfuscated or not.

Capability

Currently we can:

• Automatically scan app for vulnerabilities, reporting in HTML/JSON/text format (see below)
• Manipulate app for easier analysis: e.g. enabling debug bit, enabling full backup, disabling TLS pinning, manipulating target API level, injecting frida-gadget, etc.
• Examine app for general information
• Copy in/out app data through debug interface
• Search for certain calls/consts/sput/iput
• Deduce constants/typesets for args of op
• Scan API/private calls for native codes (NB: you need ts2-disasm-ghidra)
• Scan iOS apps for basic vulnerabilities (NB: you need ts2-disasm-ghidra)
• Attach frida scripts
• Provide frida interative session
• Tracing calls
• etc.

Installation

Containers

NOTE:

• As of 2.1.9, we are on ghcr.io. (Docker Hub is somewhat deprecated)
• Requires adbd in the host to control devices.

We provide containers so you can use right away as follows; now this is also the recommended way, and the only way if you are on Windows, to run:

$ docker run --rm -v $(pwd):/out -v ts2:/cache ghcr.io/alterakey/trueseeing

If you want to run statelessly you omit mounting volume onto /cache (not recommended for day-to-day use though; also see #254):

$ docker run --rm -v $(pwd):/out ghcr.io/alterakey/trueseeing

Install with uv

Alternatively, you can install our package with uv as follows. Especially the uv tool install form of installation might be useful for extensions (see below), as it grants them the greatest freedom. Just remember you need a JRE and Android SDK (optionally; to mess with devices):

$ uvx trueseeing

$ uv tool install trueseeing
$ trueseeing

Install with pip (deprecated)

Of course you can always use the good old pip if you must:

$ pip install trueseeing

Usage

Interactive mode

You can interactively scan/analyze/patch/etc. apps -- making it the ideal choice for manual analysis:

$ trueseeing target.apk
[+] trueseeing x.y.z
ts[target.apk]> ?
...
ts[target.apk]> i                      # show generic information
...
ts[target.apk]> pf AndroidManifest.xml # show manifest file
...
ts[target.apk]> a                      # analyze resources too
...
ts[target.apk]> /s something           # search text
...
ts[target.apk]> as                     # scan
...
[+] done, found 6403 issues (174.94 sec.)
ts[target.apk]> gh report.html

Batch mode

We accept an inline command (-c) or script file (-i) to run before giving you prompt, as well as quitting right away instead of prompting (-q; we don't require a tty in this mode!).

You can use the features to conduct a batch scan, as follows e.g. to dump findings right onto the stderr:

$ trueseeing -eqc 'as' target.apk

To generate a report file in HTML format:

$ trueseeing -eqc 'as;gh report.html' target.apk

To generate a report file in JSON format:

$ trueseeing -eqc 'as;gj report.json' target.apk

To get report generated in stdout, omit filename from final g* command:

$ trueseeing -eqc 'as;gh' target.apk > report.html
$ trueseeing -eqc 'as;gj' target.apk > report.json

Non-interactive scan mode (deprecated)

Traditionally, you can scan apps with the following command line to get findings listed in stderr:

$ trueseeing --scan target.apk

To generate a report in HTML format:

$ trueseeing --scan --scan-output report.html target.apk
$ trueseeing --scan --scan-report=html --scan-output report.html target.apk

To generate a report in JSON format:

$ trueseeing --scan --scan-report=json --scan-output report.json target.apk

To get report generated in stdout, specify '-' as filename:

$ trueseeing --scan --scan-output - target.apk > report.html
$ trueseeing --scan --scan-report=html --scan-output - target.apk > report.html
$ trueseeing --scan --scan-report=json --scan-output - target.apk > report.json

Advanced Usages

Extensions

You can write your own commands and signatures as extensions. Extensions are placed under /ext (containers) or ~/.trueseeing2/extensions/ (uv/pip) . Alternatively you can distribute your extensions as wheels. We provide type information so you can not only type-check your extensions with zuban but also get a decent assist from IDEs. See the details section for details.

Build

You can build it as follows:

$ docker build -t trueseeing https://github.com/alterakey/trueseeing.git#main

To build wheels you can do with flit, as follows:

$ flit build

To hack it, you need to create a proper build environment. With uv you could just do:

$ git clone https://github.com/alterakey/trueseeing.git wc
$ uv sync --locked --dev
$ (... hack ...)
$ uv run trueseeing ...                                    # to run
$ uv run zuban check trueseeing && uv run ruff trueseeing  # to validate
Success: no issues found in XX source files
$ uv run flit build                                        # to build (wheel)
$ docker build -t trueseeing .                             # to build (container)

With pip, to create one, firstly set up a venv, install flit and validating toolchains (zuban and ruff) in there, and have flit pull dependencies. In short, do something like this:

$ git clone https://github.com/alterakey/trueseeing.git wc
$ python3 -m venv wc/.venv
$ source wc/.venv/bin/activate
(.venv) $ pip install flit zuban ruff
(.venv) $ flit install --deps=develop -s
(.venv) $ (... hack ...)
(.venv) $ trueseeing ...                                   # to run
(.venv) $ zuban check trueseeing && ruff check trueseeing  # to validate
Success: no issues found in XX source files
(.venv) $ flit build                                       # to build (wheel)
(.venv) $ docker build -t trueseeing .                     # to build (container)

Details

Vulnerability Classes

Currently we can detect the following class of vulnerabilities, largely ones covered in OWASP Mobile Top 10 - 2016:

• Improper Platform Usage (M1)

  • Debuggable
  • Inadvent publishing of Activities, Services, ContentProviders, BroadcastReceivers

• Insecure Data (M2)

  • Backupable (i.e. suspectible to the backup attack)
  • Insecure file permissions
  • Logging

• Insecure Commnications (M3)

  • Lack of pinning (i.e. suspictible to the TLS interception attack)
  • Use of cleartext HTTP
  • Tamperable WebViews

• Insufficient Cryptography (M5)

  • Hardcoded passphrase/secret keys
  • Vernum ciphers with static keys
  • Use of the ECB mode

• Client Code Quality Issues (M7)

  • Reflectable WebViews (i.e. XSSs in such views should be escalatable to remote code executions via JS reflection)
  • Usage of insecure policy on mixed contents

• Code Tampering (M8)

  • Hardcoded certificates

• Reverse Engineering (M9)

  • Lack of obfuscation

Extension API