Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
guides/mobile-app-instrumentation-playbook.md — End-to-end operational playbook (iOS/Android runtime model, instrumentation lab build-out on jailbroken iOS / Magisk-rooted Android, instrumentation workflow, real-world case studies of Instagram cookie hijack, Snapchat jailbreak detection history, Uber cert pinning evolution, TikTok keystore bypass, and defense patterns including Play Integrity API and iOS App Attest)
guides/mobile-app-instrumentation-ios-android-deep.md — Advanced deep dive: iOS instrumentation internals (frida-server / frida-gadget / USBMUXD, SSL Kill Switch 3, BoringSSL bypass, Shadow / A-Bypass / Liberty / Choicy jailbreak detection defeat, Keychain dump, URL scheme hijack, App Attest bypass research, PAC on A12+, CoreTrust CVE-2023-41974 / kfd exploit chain), Android instrumentation (Frida gadget injection modes, Magisk + Zygisk + LSPosed, RASP bypass recipe for PGY / Promon / Arxan, Flutter reverse via reflutter + BoringSSL, React Native / Hermes via hbctool), high-value CVE targets (CVE-2024-0044 run-as sandbox escape, Play Integrity bypass via pif), the full SSL Pinning Bypass Catalog decision matrix (15+ libraries mapped to bypass technique), modern obfuscation defeat (OLLVM string deobfuscation, control-flow flattening via angr / HexRaysDeob / ghidra-emotet, radare2 native arm64 flow), and lab setup (Corellium, Android Studio Emulator + Frida, Genymotion, Objection tunneling)
Summary
Mobile app instrumentation is the runtime counterpart to static mobile-app analysis. Where mobile-security reads the Info.plist, the AndroidManifest.xml, the binary's strings, and the Smali bytecode to discover what the app could do, this skill attaches a live Frida/Objection/r2frida bridge to a running process and observes what the app actually does at execution time — then forces it to behave otherwise. The operator spins up a jailbroken iOS device (checkm8 on A7-A11 or newer arm64e on A12+ with appropriate exploits) or a Magisk-rooted Android device (Android 13/14/15), acquires the app binary (decrypted IPA on iOS via Clutch/iPAA/bfinject; APK pull on Android via ), performs static triage with /// to locate the classes and functions of interest, then attaches Frida to hook arbitrary Java/Kotlin methods, Objective-C/Swift methods, and native exported functions. With those hooks in place, the operator can dump the iOS keychain (), enumerate and manipulate the Android Keystore (), bypass SSL pinning at runtime (Frida's lineage or Objection's / ), defeat jailbreak/root detection by hooking the detection routines to return , defeat anti-debug routines (ptrace, sysctl, ), trace cryptographic operations (CC/CryptoKit on iOS; on Android), exploit WebView JavaScript bridges by invoking private bridge methods at runtime, and extract secrets (tokens, OAuth refresh tokens, biometric-gated keys) that the app would otherwise never write to disk. The discipline maps to MITRE ATT&CK Mobile T1623 (Mobile Adware), T1518 (Software Discovery), T1406 (Defacement), and T1627.001 (Adversary-in-the-Mobile-Device). Real-world engagements include bypassing Instagram certificate pinning to hijack session cookies, defeating Snapchat's evolving jailbreak detection lineage, bypassing Uber's multi-layer certificate pinning, and extracting TikTok's Keystore-attested secrets. Reference implementations: Frida (~19.7k stars, industry-standard), Objection (SensePost), r2frida (radare2 + Frida bridge), Introspy (iOS/Android tracer), Cycript (legacy iOS), iPAA/Clutch/bfinject (iOS decryption), jtool2, radare2, ghidra, jadx, apktool.
Strong/weak modes: Frida strong mode bypasses most detection (slower, harder to detect).
Gadget mode: Use frida-gadget as embedded library; no frida-server process.
Jailbreak Stealth
Root hide: Magisk DenyList; hide root from specific apps.
Systemless root: No modification to /system partition; harder to detect.
ShadowFlutter / Liberty Lite (iOS): Bypass popular jailbreak detection libraries.
SSL Pinning Bypass Stealth
Objection: Automate pinning bypass; uses Frida under hood.
Custom trust manager: Inject custom X509TrustManager; allows any cert.
Network proxy with pinning bypass: Burp + mobile companion app; bypasses pinning transparently.
Differentiation
This skill is explicitly distinct from mobile-security. Both deal with iOS/Android application assessment, but the operating mode is completely different.
Live process memory, method invocations, native calls
Tooling
jadx, apktool, mob-sf, iGoat, static analyzers
frida, objection, r2frida, introspy
What it proves
What the app could do (potential weakness)
What the app actually does at runtime (live exploitation)
Typical finding
"Hardcoded API key in strings.xml" or "Cleartext traffic allowed"
"Bypassed SSL pinning and exfiltrated the session cookie in real time"
Device requirement
None — analysis is offline
Jailbroken iOS / rooted Android device
Defender signal
Build-time config / lint
Runtime self-checks, SafetyNet, App Attest
MITRE mapping
MASVS static violations
T1627.001 — Adversary-in-the-Mobile-Device
Coexistence model: The two skills compose. A typical engagement starts with mobile-security static triage to identify candidates (which classes handle SSL pinning? which native lib implements anti-debug?), then transitions to this skill to actually bypass and exploit those candidates at runtime. Use mobile-security to scope; use this skill to demonstrate.
Difference from reverse-engineering: reverse-engineering covers the binary-reversing discipline in general (ELF, PE, Mach-O, JVM bytecode). This skill is the mobile-runtime specialization — we lean on jadx/ghidra/radare2 for static triage, but the deliverable here is a runtime finding produced by an instrumented hook, not a static disassembly.
Difference from network-sniffing-mitm: That skill intercepts wire traffic with mitmproxy/Burp as a network element. When the app implements certificate pinning, that interception fails — and that is precisely when you pivot to this skill: instrument the app to disable pinning at runtime, then network-sniffing-mitm resumes. The skills form the standard two-step attack chain against any pinned mobile backend.
Use Cases
SSL pinning bypass for MITM interception: When a target mobile app pins its backend TLS certificate (the modern default for banking, social, and enterprise apps), network-layer interception via mitmproxy/Burp fails with a TLS handshake error. Instrument the app at runtime to disable pinning — either with a generic Frida script that hooks all known pinning implementations (AFSecurityPolicy, NSURLSessionDelegate, TrustKit on iOS; OkHttp CertificatePinner, X509TrustManager, Network Security Config on Android) or with the Objection one-liner android sslpinning disable / ios sslpinning disable. Pairs directly with network-sniffing-mitm.
Jailbreak / root detection bypass: Modern apps refuse to run (or silently degrade functionality) if they detect a jailbreak or root. Instrument the detection routines — typically fcntl(path, F_GETPATH) checks against /Applications/Cydia.app, fork() probes, dyld env-var checks on iOS; RootBeer, /system writable checks, su binary probes, ro.debuggable, Magisk Hide/Zygisk detection on Android — and force them to return the clean-device answer. This is the entry gate to all other instrumentation on a real device.
In-app crypto tracing: Trace every cryptographic primitive the app invokes — CCCrypt, CC_MD5, CCHmac, CryptoKit.SHA256 on iOS; javax.crypto.Cipher, javax.crypto.Mac, java.security.MessageDigest on Android — and dump the plaintext, key, IV, and ciphertext for every operation. Recovers protocol-level secrets (signed-request HMACs, encrypted local-storage blobs, DRMs) without needing the source.
iOS Keychain extraction: Use Objection's ios keychain dump or keychain-dumper to enumerate every keychain item the app has stored, including OAuth refresh tokens, biometric-gated keys (if kSecAttrAccessibleWhenUnlocked rather than kSecAttrAccessibleWhenPasscodeSetThisDeviceOnly), and proprietary session secrets. Pairs with ios nsuserdefaults get and ios nsurlsession grab for a complete at-rest-and-in-flight secret inventory.
Android Keystore manipulation: Hook android.security.keystore and java.security.KeyStore to inspect entries, force key export (when the app does not enforce hardware attestation), bypass biometric gates (force BiometricPrompt callbacks to return AUTHENTICATION_SUCCESSFUL), and trace every signing/decryption operation. Used to recover keys protected only by software attestation and to demonstrate biometric bypasses.
Native library reversing via r2frida: For apps that offload secrets or anti-debug logic to a native .so / .dylib (common in DRM, mobile-game telemetry, and ad-fraud SDKs), attach r2frida to enumerate exports (iE), seek into native memory (s), decompile (pdg), and place Frida Interceptor.attach hooks on exported symbols. Bridges the Java/Kotlin surface and the C/C++ surface into a single live session.
Anti-debug / anti-tamper bypass: Apps that call ptrace(PT_DENY_ATTACH) on iOS or check /proc/self/status for TracerPid on Android refuse to run under Frida or a debugger. Hook ptrace to return 0, hook sysctl to hide the parent PID, hook the TracerPid string read to return 0. The standard prelude to any deeper instrumentation on hardened apps.
Runtime API hooking for arbitrary methods: The primitive that underlies everything else — pick any Java class (Java.use("com.example.LoginManager").login.implementation = ...), any Objective-C class (ObjC.classes.LoginController["- login:"](args[2])), any native export (Module.findExportByName(null, "open")) and replace its implementation, log arguments and return values, or replay with mutated inputs. Enables every other use case above.
Core Tools
Tool
Purpose
Command / Usage
Frida
Industry-standard dynamic instrumentation toolkit (JavaScript bridge to a live process on iOS/Android/Windows/Linux/macOS). ~19.7k stars.
Objective-C++ introspection runtime — injects into a process and lets you evaluate ObjC expressions live. Predates Frida on iOS; still useful on legacy iOS < 13.
cycript -p SpringBoard then UIApp.keyWindow.recursiveDescription().toString()
iPAA (iOS App Extractor)
iOS app decryption / IPA extraction — pulls the FairPlay-decrypted binary from memory on a jailbroken device.
Android APK decode/rebuild — extracts AndroidManifest.xml, smali, and res/; used to patch apps pre-instrumentation (e.g., to add a debuggable=true flag).
apktool d app.apk -o out/; apktool b out/ -o patched.apk
Methodology
Phase 1 — App Acquisition
Before any instrumentation, the operator needs the app binary on disk:
iOS: The App Store delivers FairPlay-encrypted binaries. On a jailbroken device, use Clutch -d <bundle-id> or iPAA to dump the decrypted binary from memory into a usable .app bundle. For development-signed apps (TestFlight, sideloaded), no decryption is needed. Document the iOS version (iOS 16/17/18), jailbreak used (palera1n / checkra1n / Dopamine), and whether the app uses arm64e.
Android: adb shell pm path com.example.app returns the path to the base.apk + any split APKs. Pull each with adb pull. For obfuscated/packed apps (Bangcle, Jiagu, DEX protector), dump from memory at runtime or use frida-dexdump.
Phase 2 — Environment Preparation
iOS device: Jailbroken (checkm8 for A7-A11 via palera1n/checkra1n; arm64e via Dopamine on A12-A17 / iOS 15-18). Install frida-server matching the device architecture; confirm with frida-ps -U (should list device processes).
Android device: Magisk-rooted (or LineageOS rooted); frida-server started as root in background. Android 13/14/15 may also need Zygisk + LSPosed for in-process injection. Confirm with frida-ps -U.
Anti-debug — search for ptrace, sysctl, TracerPid, debug-trace.
Crypto primitives — search for CCCrypt, Cipher.getInstance, MessageDigest.
Native libraries — list .so / .dylib in lib/, note JNI exports.
Triage output is a target list — not yet findings, but a hook candidate list.
Phase 4 — Instrumentation
Attach Frida (or Objection) and load the hook candidates:
frida -U -f com.example.app -l hook.js --no-pause for spawn-and-attach.
objection -g com.example.app explore for the REPL workflow with high-level commands.
r2 'frida://spawn/com.example.app' for native-symbol-driven work.
Iterate over candidates: pinning, jailbreak, anti-debug, crypto, keychain. Each hook logs to a file (Frida send() + Python sink) for later replay and evidence capture.
Phase 5 — Runtime Manipulation
With hooks in place, manipulate the live process:
Disable pinning → re-launch mitmproxy and capture the TLS-decrypted traffic.
Bypass jailbreak → trigger gated features (e.g., banking transfers on jailbroken iOS).
Force crypto keys → replay requests offline with extracted HMAC/AES keys.
Replay JWTs extracted from the keychain against the backend with curl for a clean client-to-server compromise chain.
Document every step: hook script, log output, screenshot, and the downstream consequence (decrypted HTTPS, extracted token, bypassed gate). This evidence packet is the deliverable.
iOS Instrumentation Stack
The iOS instrumentation stack spans multiple layers — the operator must know which layer a given app uses to choose the right bypass. This stack is documented in detail in guides/mobile-app-instrumentation-ios-android-deep.md.
Frida runtime variants: frida-server (daemon, requires jailbreak, exposes 127.0.0.1:27042 via USBMUXD); frida-gadget (FridaGadget.dylib patched into the app's .app bundle and re-signed — works on non-jailbroken devices and evades frida-server detection). On A12+ arm64e devices, only the arm64e build of frida-server / frida-gadget can attach to native arm64e apps; PAC (Pointer Authentication Codes) forces this architecture match.
SSL Kill Switch 3 (https://github.com/nicklama/ssl-kill-switch3) — system-wide pinning bypass at SecTrustEvaluate + NSURLSession. Lowest-effort bypass for iOS 15-17, but does NOT cover BoringSSL, custom URLSessionDelegate, or Network.framework-based pins. Install via Chariz repo on Dopamine jailbreak.
BoringSSL / libssl bypass — Flutter iOS apps and any app using cronet/grpc ship a static-linked BoringSSL. Bypass via Interceptor.replace on SSL_set_custom_verify / SSL_CTX_set_custom_verify / SSL_verify_peer_chain after pattern-scanning libflutter.so for the BoringSSL API pointer table.
objection advanced patterns — ios hooking set return_value "-[SecurityManager isJailbroken]" false is the most powerful jailbreak-detection bypass primitive; ios hooking watch class UrlSessionDelegate enumerates a class's full method surface; ios bundles list_frameworks confirms whether TrustKit, AFNetworking, or CryptoKit is linked.
Shadow (opa334) — full environment simulation; the operator's default for hardened apps. Replaces the entire userspace environment the app sees (hides /Applications, blocks dyld injection, fakes sysctl returns, scrubs task_get_exception_ports).
A-Bypass (repo.co.kr, paid) — hooks every known JB-detection syscall (fork, stat, access, lstat, open, dlopen, sysctl, ptrace, task_get_exception_ports, dyld_image_count, getppid).
Liberty Lite / Liberty Duo (level3tjg.xyz, free) — covers iOS 13-16 banking apps.
Choicy (opa334) — selectively disable tweaks per-app; many banking apps detect any injected tweak, so you disable everything except Shadow/A-Bypass for the target app.
iOS high-value targets — Keychain dump (keychain-dumper root-run via SecItemCopyMatching or objection's ios keychain dump), URL scheme hijack (Info.plistCFBundleURLTypes + openURL validation), App Attest partial bypass (replay / dev-tier acceptance — the cryptographic chain is NOT bypassable as of 2026), PAC constraints on A12+ (use arm64e builds of frida-server), CoreTrust bypass via CVE-2023-41974 / kfd exploit chain (iOS 16.5-16.6.1 only).
Android Instrumentation Stack
The Android stack has its own dynamics — gadget injection for non-rooted devices, Magisk + Zygisk + LSPosed as a Frida-invisible alternative, and a separate RASP-bypass discipline.
Frida gadget injection modes — patch libfrida-gadget.so into lib/<abi>/ of a decompiled APK via apktool + LIEF, configure via libfrida-gadget.config.so:
"listen" — binds 127.0.0.1:27042, connect via frida -H after adb forward. Default, most flexible.
"script" — auto-loads libfrida-gadget.script.so (a JS file with .so suffix to fool APK packers). Stealth: no listening port.
"connect" — connects out to a remote frida-server. Useful for device-farm labs.
Gadget injection is the operator's default for non-rooted engagements (corporate-managed devices) and for stealth against RASP that detects frida-server.
Magisk + Zygisk + LSPosed — modern root stack. Magisk (root + Zygisk injection); Zygisk "Enforce DenyList" hides root from specified apps; LSPosed (modern Xposed) provides a Frida-invisible Java/Kotlin hooking path. Many RASP products (PGY, Promon) detect Frida by port/process/maps but not Xposed/LSPosed — when Frida is detected, drop to LSPosed module development.
Port-shift Frida: ./frida-server -l 0.0.0.0:13371 (defeats the 27042 port check).
Rename the binary (defeats /proc/<pid>/exe name check).
Scrub /proc/self/maps and /proc/self/status via fopen interceptor (defeats the frida-agent / libfrida maps scan).
Rename Frida threads via pthread_setname_np interceptor (defeats the gmain / gdbus / gum-js-loop thread-name scan).
Hook Debug.isDebuggerConnected to return false (defeats the debugger check).
Android high-value targets — KeyStore abuse (CVE-2024-0044 run-as sandbox escape on Android 12-13), Play Integrity bypass via pif (PlayIntegrityFix) Magisk module + MagiskHidePropsConf for MEETS_DEVICE_INTEGRITY (no public bypass for MEETS_STRONG_INTEGRITY), content provider confusion (Drozer enumeration + Frida query logging), Intent redirection patterns (Frida hook on startActivity to detect nested Intents — CVE-2020-7663 Slack lineage).
Flutter app reverse — reflutter (https://github.com/Impact-I/reflutter) patches the APK to disable BoringSSL verify and route through mitmproxy; frida-dexdump dumps DEX; doldrums / Blutter parsers recover Dart class hierarchy from _kDartIsolateSnapshotInstructions. Standard OkHttp / X509TrustManager bypass scripts do NOT work on Flutter apps because no Java network layer is involved — must bypass BoringSSL inside libflutter.so.
React Native reverse — apktool d extracts assets/index.android.bundle; for Hermes bytecode (React Native 0.70+) use hbctool disasm to recover the instruction stream and string table. Runtime: hook com.facebook.hermes.reactexecutor.HermesExecutor.callJSCallback to log every JS → native call.
SSL Pinning Bypass Catalog
Modern mobile apps pin TLS at multiple layers. The decision matrix below maps each pinning library to its bypass technique. Use it as the first lookup when triaging an app.
Library / Mechanism
Platform
Detection Signature
Bypass Technique
OkHttp CertificatePinner
Android
okhttp3.CertificatePinner.check
Objection android sslpinning disable or hook check() to no-op
X509TrustManager (custom)
Android
Class implements javax.net.ssl.X509TrustManager
Replace with trust-all TrustManager via Java.registerClass
Operator workflow: identify the pinning library (search the disassembled app), match against the matrix, apply the bypass technique, verify via mitmproxy. Modern banking apps stack 2-4 pinning layers (HTTP client + OS + native SDK) — the operator must bypass all layers; bypass scripts compose without conflict. See guides/mobile-app-instrumentation-ios-android-deep.md Section 6 for the full decision matrix and the stacked-pinning case study.
Modern Obfuscation Defeat
Modern mobile native libraries (banking, gaming, DRM, anti-cheat) are protected by OLLVM (Obfuscator-LLVM) descendants. The same techniques that appear in desktop malware (Emotet, Conti, TrickBot) appear in mobile libNative.so / Native.dylib.
OLLVM string deobfuscation — encrypted strings with per-string keys, decrypted in a global constructor. Detect via Ghidra/IDA: functions with a long XOR loop over .bss. Bypass strategies: runtime dump via Frida Memory.scan of .bss after app launch; IDA Python / Ghidra script that pattern-matches the XOR-decryptor; hook the decryption function (sub_XXXX) to log decrypted plaintext.
Control-flow flattening (CFF) — function body reassembled under a switch dispatcher with a state variable. Detect in IDA graph view: "bicycle wheel" shape. Bypass strategies: symbolic execution via angr recovers the original CFG; HexRaysDeob IDA plugin (Rolf Rolles) transforms CFF back to structured form in the decompiler; ghidra_emotet Akamai script auto-reconstructs any CFF'd function.
IDA Python / Ghidra scripts — for IDA Pro: identify large functions (>0x400 bytes), count cmp+jcc patterns to flag dispatcher candidates. For Ghidra: use the Jython interface to scan for XOR instructions in long functions (string-decryptor heuristic).
Native arm64 reverse via radare2 — r2 -A libNative.so; afl~Java_ to list JNI exports; pdf @ sym.Java_com_example_Native_nativeSign to disassemble; axt @ <addr> for cross-references; /x for pattern scanning. Bridge to live analysis via =+frida://spawn/<pkg>.
Virtualization-based protections (Themis, VMProtect mobile, Inside Secure) — the function is compiled into a custom bytecode interpreter; full reversal requires identifying the interpreter's opcode table and writing a devirtualizer. Research-grade; outside the scope of a typical engagement.
See guides/mobile-app-instrumentation-ios-android-deep.md Section 7 for full worked examples (Frida .bss dump script, angr CFF recovery, IDA Python dispatcher finder, radare2 JNI flow).
Practical Steps
Example 1 — Frida SSL Pinning Bypass (Generic Script)
Save as bypass_ssl_pinning.js:
// bypass_ssl_pinning.js
// Generic SSL pinning bypass covering common iOS and Android pinning APIs.
// Usage: frida -U -f com.example.app -l bypass_ssl_pinning.js --no-pause
setTimeout(function () {
// ---- Android: OkHttp CertificatePinner ----
if (Java.available) {
Java.perform(function () {
try {
var CertificatePinner = Java.use('okhttp3.CertificatePinner');
CertificatePinner.check.overload('java.lang.String', 'java.util.List').implementation = function () {
console.log('[+] OkHttp CertificatePinner.check() bypassed');
return;
};
} catch (e) { console.log('[-] OkHttp not present: ' + e); }
// ---- Android: X509TrustManager ----
try {
var X509TrustManager = Java.use('javax.net.ssl.X509TrustManager');
var SSLContext = Java.use('javax.net.ssl.SSLContext');
var TrustManager = Java.registerClass({
name: 'org.owasp.TrustAll',
implements: [X509TrustManager],
methods: {
checkClientTrusted: function () {},
checkServerTrusted: function () {},
getAcceptedIssuers: function () { return []; }
}
});
SSLContext.init.overload(
'[Ljavax.net.ssl.KeyManager;',
'[Ljavax.net.ssl.TrustManager;',
'java.security.SecureRandom'
).implementation = function (km, tm, sr) {
console.log('[+] SSLContext.init() replaced with trust-all');
this.init(km, [TrustManager.$new()], sr);
};
} catch (e) { console.log('[-] X509TrustManager hook failed: ' + e); }
});
}
// ---- iOS: SecTrustEvaluate ----
if (ObjC.available) {
var SecTrustEvaluate = Module.findExportByName('Security', 'SecTrustEvaluate');
if (SecTrustEvaluate) {
Interceptor.replace(SecTrustEvaluate, new NativeCallback(function (trust, result) {
// Force kSecTrustResultProceed = 1
Memory.writePointer(result, ptr(1));
console.log('[+] SecTrustEvaluate() forced to proceed');
return 0; // errSecSuccess
}, 'int', ['pointer', 'pointer']));
}
// Also hook the Swift-level URLSessionDelegate if present (left as exercise).
}
}, 0);
Run with frida -U -f com.example.app -l bypass_ssl_pinning.js --no-pause. All TLS errors should disappear — re-arm mitmproxy and the traffic flows decrypted.
Example 2 — Objection Runtime Exploration
# Spawn the target under objection's REPL
objection -g com.example.app explore
# Inside the REPL — explore the app surface:
android hooking list classes
android hooking search classes keyword
android hooking list activities
android hooking list services
android hooking watch class_method com.example.LoginManager.login --dump-args --dump-return
# Disable SSL pinning in-process
android sslpinning disable
ios sslpinning disable
# Dump secrets at rest
ios keychain dump
ios nsuserdefaults get
android keystore list
android root disable
# Bypass biometrics
android biometric disable
Objection is the fastest path from "I have the app" to "I have the runtime surface" — use it for triage, drop down to raw Frida for custom hooks.
Example 3 — r2frida Native Library Analysis
# Launch r2 against the live process
r2 'frida://spawn/com.example.app'
# Inside r2 — enumerate native modules:
[0x00000000]> iE~libnative
# Seek into the native lib and list exports:
[0x00000000]> s sym.Java_com_example_Native_nativeSign
[0xdeadbeef]> pdf # disassemble
[0xdeadbeef]> =?/ip # inject a Frida hook for this export via the bridge
# Detach and resume:
[0x00000000]> dc
r2frida lets you do static and dynamic analysis in one session — invaluable for understanding a native routine before you commit to a Frida hook.
Example 4 — Android Keystore Hook for Crypto Tracing
// trace_keystore.js — log every Cipher.init + doFinal with key + plaintext
Java.perform(function () {
var Cipher = Java.use('javax.crypto.Cipher');
Cipher.init.overload('int', 'java.security.Key').implementation = function (mode, key) {
console.log('[Cipher.init] mode=' + mode + ' key=' + key.getEncoded());
return this.init(mode, key);
};
Cipher.doFinal.overload('[B').implementation = function (input) {
var out = this.doFinal(input);
console.log('[Cipher.doFinal] in=' + bytesToHex(input) + ' out=' + bytesToHex(out));
return out;
};
function bytesToHex(arr) {
var s = '';
for (var i = 0; i < arr.length; i++) s += ('0' + (arr[i] & 0xff).toString(16)).slice(-2);
return s;
}
});
Defense Perspective
Defenders should assume any non-server-side check is bypassable with sufficient time. The mitigation strategy is layered runtime defenses and server-side attestation:
iOS App Attest / DeviceCheck: Server validates an Apple-signed attestation that the request originated from a genuine app instance on a non-jailbroken device. Defeats runtime instrumented replays. Pair with per-session server nonce binding to prevent assertion replay; reject dev-tier attestations in production.
Android Play Integrity API (successor to SafetyNet): Server validates a Google-signed integrity verdict covering app authenticity, device integrity (CTS-passing, no known root), and account licensing. Demand MEETS_STRONG_INTEGRITY for high-value operations (no public bypass exists as of 2026); MEETS_DEVICE_INTEGRITY is the realistic baseline but is bypassable via pif / Magisk Hide.
Multi-layer jailbreak / root detection: Combine fcntl/dyld/fork checks on iOS and RootBeer + /proc/self/maps + /system/bin/su + Magisk Hide detection on Android. Detect Frida by scanning /proc/self/maps (Android) or dyld_image_count (iOS) for frida-agent / libfrida. Be aware of the bypass recipe (port-shift, binary rename, /proc/self/maps scrub, thread-rename, Debug.isDebuggerConnected hook) and ship counter-measures for each: monitor for renamed frida processes, scrub-resistant maps via readlink("/proc/self/exe") cross-checks, additional checks for Zygisk modules.
Runtime self-integrity checks: Periodically checksum the app's .text section and detect Frida stalker / interceptor trampolines by walking dyld / dlopen tables. Detect objection by scanning for the FRIDA_SCRIPT_AGENT symbol or unusual Java.use patterns in ART.
Hardware-attested keys: On Android, generate keys inside the hardware-backed Keystore (setUserAuthenticationRequired(true) + setAttestationChallenge) so even a hooked Cipher cannot export them. Verify the attestation certificate chain server-side against the Google Hardware Attestation root. On iOS, generate attestation keys via DCAppAttestService so keys live in the Secure Enclave.
Certificate pinning with multiple pins + fallback + native-layer enforcement: Pin the leaf, intermediate, and CA; ship a pin rotation channel; have a documented fail-closed policy (the app must abort on pin failure rather than degrade to system trust). Stack pinning at multiple layers — HTTP client (OkHttp CertificatePinner / AFSecurityPolicy), OS (X509TrustManager / SecTrustEvaluate), and native (BoringSSL SSL_set_custom_verify inside a packed libNative.so with OLLVM control-flow flattening). Each layer raises bypass cost independently.
Mobile app hardening (build-time + runtime):
Obfuscation — apply R8/ProGuard (Android) and Swift symbol stripping + LLVM obfuscation (iOS) to slow static triage. Apply OLLVM control-flow flattening to native libraries holding secrets.
Anti-tamper — sign the APK with v2/v3 scheme; verify the signature at app startup and inside native code (detect apktool rebuilds that resign with a different cert). On iOS, embed dlsym checks for non-Apple frameworks and abort on unknown dylib load.
Anti-debug — call ptrace(PT_DENY_ATTACH) on iOS (defeats lldb; bypassable but raises the bar); check /proc/self/statusTracerPid on Android; integrate a commercial anti-debug SDK (Promon, Arxan, Guardsquare) for high-value targets.
RASP — Runtime Application Self-Protection SDKs (Promon SHIELD, Guardsquare DexGuard, Verimatrix, PGY/Pangu) combine anti-tamper, anti-debug, anti-Frida, root/JB detection, and method-swizzling protection into a single vendor library. Be realistic about their limits — every commercial RASP has been bypassed publicly within 6-12 months of release.
String encryption — encrypt API keys, OAuth client secrets, and HMAC keys in the binary; decrypt only inside native code at runtime. Slows Frida strings-based secret extraction.
Screenshot / screen-recording protection — flag sensitive screens with FLAG_SECURE (Android) / isProtected (iOS) to block screenshot/Mirroring-based exfiltration.
Tapjacking / overlay protection — reject touches when an overlay is detected (Android filterTouchesWhenObscured).
Defense-in-depth logging: Report integrity failures back to the server; rate-limit or revoke compromised sessions even if the client-side gate is bypassed. Correlate client integrity signals (App Attest / Play Integrity verdict, device fingerprint, behavioral biometrics) for high-value operations.
Server-side rate-limiting and anomaly detection: Client-side checks always fail eventually; the server must detect anomalous patterns (a single user making thousands of API calls per minute, requests from unusual geolocations, replayed nonces) and revoke.
A thorough mobile-app red-team deliverable should be paired with a mitigation mapping table: each finding → corresponding MASVS-RESILIENCE control → recommended runtime defense. See guides/mobile-app-instrumentation-playbook.md Section "Defense Patterns" for the full mapping.
