iOS Secure Development Guide

iOS Security

This guide is a collection of the most common vulnerabilities found in iOS applications. The focus is on vulnerabilities in the applications’ code and only marginally covers general iOS system security, Darwin security, C/ObjC/C++ memory safety, or high-level application security.

Nevertheless, hopefully the guide can serve as training material to iOS app developers that want to make sure that they ship a more secure app. Also, iOS security reviewers can use it as a reference during assessments.

Just like any software, this guide will rot unless we update it. We encourage everyone to help us on that, just open an issue or send a pull request!

API-level issues

API: Generate cryptographically-strong random numbers

Generally, iOS provides easy-to-use cryptographic interfaces. Don’t implement custom crypto algorithms (besides crypto problems, it can also cause issues during App Store review).

Only supply cryptographically-strong random numbers to cryptographic functions.

Audit tip: Check that all cryptographically secure random numbers are fetched using the Randomization Services programming interface.

Correct example:

int r = SecRandomCopyBytes(kSecRandomDefault, sizeof(int), (uint8_t*) &res);

API: Prevent leaking sensitive data during app backgrounding

When iOS backgrounds an app, a screenshot of the app used to get saved to an unencrypted cache on the local file system. This happens for example when the user presses the home button. Apple recommends developers to hide any sensitive information before this occurs. However, when testing iOS 10, the screenshot is stored in the encrypted app sandbox. Therefore it is less of a risk.

If the app is handling sensitive user data, verify that code exists to hide or blur the sensitive elements or the full window.

Audit tip: Check for hiding code in applicationDidEnterBackground.

Alternatively, you can set allowScreenShot. Using ignoreSnapshotOnNextApplicationLaunch seems broken.

API: Handle the pasteboard securely

If the pasteboard is marked persistent it may get saved to local storage along with potentially sensitive user data. Also, make sure to clear pasteboard when an application backgrounds.

Audit tip: Check for UIPasteboardNameGeneral & UIPasteboardNameFind.

API: Disable auto-correction for sensitive input fields

Some iOS versions cache keyboard entries for auto-correction. This is disabled for password fields but should be disabled for other sensitive fields (e.g. credit card number) as well. Set the following to prevent this:

UITextField autoCorrectionType = UITextAutocorrectionTypeNo

or mark the text field as secure (hidden input) with the secureTextEntry attribute.

Audit tip: Check for sensitive non-password input fields (e.g. credit card) which do not have UITextAutoCorrectionNo.

Data-handling issues

Handling data: Deserialize data securely

During deserialization, some objects are re-instantiated in memory. Thus, if the serialized data originates from an untrusted source, code execution might be possible.

When writing your own classes, it is generally a good idea to comply with the NSSecureCoding protocol, to make sure that classes constructed from external sources are the intended class. It is also required by Apple for classes that are used with inter-application communication (UIActivityViewController).

Audit tip: Check for insecure deserialization from untrusted sources. Some deserialization (NSCoding, NSCoder) must have checks for the deserialized data to be within bounds.

Audit tip: Other deserialization (CFBundle, NSBundle, NSKeyedUnarchiverDelegate, didDecodeObject, awakeAfterUsingCoder) can directly lead to code execution by returning different objects during deserialization.

Audit tip: Check that nib files are not dynamically loaded from untrusted origins.

Handling data: Avoid SQL Injection

If attacker-supplied strings are concatenated to a SQL query, SQL injection on a sqlite database may occur. This might leak sensitive information from the database or inject malicious payloads.

Audit tip: Check for calls to sqlite3_exec() and other non-prepared SQL functions. The functions sqlite3_prepare*() should be used instead.

Incorrect example:

NSString *uid = [myHTTPConnection getUID];
NSString *statement = [NSString StringWithFormat:@"SELECT username FROM users
where uid = '%@'",uid];
const char *sql = [statement UTF8String];

Correct example:

const char *sql = "SELECT username FROM users where uid = ?";
sqlite3_prepare_v2(db, sql, -1, &selectUid, NULL);
sqlite3_bind_int(selectUid, 1, uid);
int status = sqlite3_step(selectUid);

What's even worse, libsqlite3.dylib in iOS supports fts3_tokenizer function, which has two security issues by design. This SQL function has two prototype:

SELECT fts3_tokenizer(<tokenizer-name>);
SELECT fts3_tokenizer(<tokenizer-name>, <sqlite3_tokenizer_module ptr>);

The first from can be abused to leak the base address of libsqlite3.dylib, which breaks ASLR.

FMResultSet *s = [db executeQuery:@"SELECT hex(fts3_tokenizer('simple')) as fts;"];
while ([s next]) {
    NSString *val = [s stringForColumn:@"fts"];
    NSLog(@"val: %@", val); // the address of simpleTokenizerModule in libsqlite3.dylib, in big endian
}

If the second argument is given, it registers a new tokenizer and the argument is the address of a virtual function table. This will lead to native code execution via SQLite3 callbacks:

[db executeUpdate:@"select fts3_tokenizer('simple', x'4141414141414141');"]; // a fake virtual table
[db executeUpdate:@"drop table a if exists;"]; // in case the virtual table already extst
FMResultSet *result = [db executeQuery:@"create virtual table a using fts3;"];
NSLog(@"%d", [result next]); // trigger pointer dereference

The crash information:

thread #1: tid = 0x19ac77, 0x0000000184530764 libsqlite3.dylib`___lldb_unnamed_symbol1073$$libsqlite3.dylib + 1500, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x4141414141414149)

App hardening

Hardening: Enable exploit mitigation compile-time options

In order to make exploiting iOS applications harder for the attacker, make sure you enable platform exploit mitigation options.

Audit tip: Check that compiler and linker flags for exploit mitigation are enabled.

Flags to enable:

Objective-C automatic reference counting (-fobjc-arc) helps to prevent use-after-free and use-after-release bugs. Enabling ARC might not be always possible with shared code, performance-sensitive code, or legacy codebases. Check with:
otool -I -v binary | grep _objc_autorelease
Stack smashing protection (-fstack-protector-all). This potentially helps to prevent stack buffer overflows. Check with (should be on by default):
otool -I -v binary | grep stack_chk_guard
Full ASLR - position independent executable (-pie). This makes it harder for the attacker to find known code locations. (Apple App Store guards this for iPhone 5+ targets). Check with (should be on by default):
otool -hv binary | grep PIE

Hardening: Check Xcode’s static analysis report

Static analysis can help to reveal memory leak, use-after-free, use-after-release, and other bugs.

Audit tip: Check the output of Xcode’s "Build & Analyze"

Hardening: Check that support for third party keyboards is disabled

By default, iOS8+ allows third-party applications to override the built-in keyboard with may leak keystrokes or words to untrusted parties. Depending on the application risk profile, this may be both a Security and Compliance issue. This is how it can be disabled in Swift.

Network-level issues

Networking: Use GTMSessionFetcher communication securely

By default, GTMSessionFetcher won’t load any non-https URL schemes.

Audit tip: Check that no exceptions are made by using allowedInsecureSchemes, allowLocalhostRequest or GTM_ALLOW_INSECURE_REQUESTS.

Networking: Configure App Transport Security (ATS)

By default, apps linked against iOS 9 cannot make unprotected HTTP connections. Review that the ATS configuration is correct.

Audit tip: Check that no exceptions are done in the Info.plist.

Audit tip: Check that the list of HTTPS domains in the Info.plist is correct.

In iOS 10, some new exceptions are available:

Exception for streaming media using AVFoundation
NSAllowsArbitraryLoadsInWebContent will exempt ATS in WKWebView

Networking: Use native TLS/SSL securely

SSL should be used on all communication to prevent attackers from reading or modifying traffic on the network.

Audit tip: Check that all APIs besides local WebViews use SSL (https scheme, no http).

Audit tip: Check that authorization tokens are never be passed in URLs but only in headers of HTTPS requests (e.g. as a Cookie header). The concern here is that they are unintentionally logged on a ISP/company proxy or accidentally leaked through referrers without the user’s knowledge.
Audit tip: Check that no debug options for SSL have been enabled in the release build:
NSStream: kCFStreamSSLLevel kCFStreamSSLAllowsExpiredCertificates kCFStreamSSLAllowsAnyRoot kCFStreamSSLAllowsExpiredRoots kCFStreamSSLValidatesCertificateChain NSURLRequest setAllowsAnyHTTPSCertificate NSURLConnection continueWithoutCredentialForAuthenticationChallenge ValidatesSecureCertificate setValidatesSecureCertificate

Issues with IO

IO: Validate incoming URL handler calls

URI handlers are special entry points to the application and can be called from email, chat, browser or other applications. They can be used as delivery vehicles for attacks that exploit logic bugs, XSS, XSRF-style bugs or buffer-overflows.

Audit tip: Check for URI handlers registered and handled by the application (registerForRemoteNotificationTypes and handleOpenURL).

To illustrate the problem, some attack ideas that could be feasible:

myapp://cmd/run?program=/path/to/program/to/run
myapp://cmd/set_preference?use_ssl=false
myapp://cmd/sendfile?to=evil@attacker.com&file=some/data/file
myapp://cmd/delete?data_to_delete=my_document_ive_been_working_on
myapp://cmd/login_to?server_to_send_credentials=malicious.webserver.com
myapp://cmd/adduser='>"><script>javascript to run goes here</script>
myapp://use_template?template=/../../../../../../../../some/other/file

Audit tip: Check that userInfo and launchOptions are validated during parsing of URI request. For actions after the URL handler, it is important to ask user for confirmation before taking action.

Additionally, note that other applications could be able to register the same URL handler and intercept requests. When passing highly sensitive information it is preferable to sign and/or encrypt URL handler-transmitted data to prevent leakage and/or forgery.

IO: Validate outgoing requests and URL handlers

Audit tip: Check for outgoing requests made by an UIWebView. Only a certain whitelist of schemes should be allowed (http/https) to avoid file:, facetime:, facetime-audio:, sms:, or other app-id: URLs. Make sure to filter tel: URLs (or require user confirmation) because they can be used to automatically dial a cost incurring phone number.

The correct way to check an outgoing request is shown below:

- (BOOL)webView:(UIWebView *)webView shouldStartLoadWithRequest:(NSURLRequest
  *)request navigationType:(UIWebViewNavigationType)navigationType;

If you are using WKWebView you'll need to use the -webView:decidePolicyForNavigationAction:decisionHandler: method in the WKNavigationDelegate protocol to catch requests like these.

IO: Prevent WebView UI redressing

Audit tip: Check for WebViews which would allow for browser UI redressing, for example a full screen WebView which could display an UI similar to the original App or to a login screen. Such WebViews could be used by attackers to do phishing.

Audit tip: Check for WebViews which would allow browsing the web like a browser but don't provide typical browser security UI like an URL bar indicating the domain and TLS status. Also, make sure that if the WebView allows browsing the web, common browser security feature like mixed-content prevention are still present in the WebView.

IO: Avoid XSS in WebViews

Audit tip: Check how the UIWebView/WKWebView is handling strings because attacks similar to XSS can occur. An XSS in a UIWebView can potentially leak local files, for example the address book and cookies. XSS in WKWebView is more restricted because AllowUniversalAccessFromFileURLs and AllowFileAccessFromFileURLs are off by default. Also make sure that the WebView is not prone to redirection which can be utilized for phishing.

IO: Avoid local HTML preview with UIWebView

Audio tip: Check if file preview functionality is implement with UIWebView. It has the same impact with XSS, except the whole page is under control of attackers. Since the origin is file://, UIWebView allows read local files and send AJAX request to arbitrary third party websites.
Make sure to use QLPreviewController to preview file attachments. It disables javascript on iOS <=9, otherwise it uses WKWebView which doesn't allow local file and cross domain internet access by default.

Memory corruption issues

Memory: Prevent NULL byte injection

CF/NS strings contain NULL bytes at different locations. When an insecure conversion occurs a string could terminate early.

Audit tip: Check for incorrect conversion between the raw bytes of the CFDataRef / CFStringRef / NSString and C strings.

This example shows incorrect conversion:

NSString *fname = @"user_supplied_image_name\0";
NSString *sourcePath = [[NSString alloc] initWithFormat:@"%@/%@.jpg",
                        [[NSBundle mainBundle] resourcePath],
                        fname];
printf("%s", [sourcePath
UTF8String]);
// prints [...]Products/Debug/user_supplied_image_name without the .jpg ending

Memory: Prevent format string attacks

Format string attacks can be mounted on traditional functions (printf, scanf, syslog, etc.), but also on iOS platform functions. The Xcode Build & Analyze option should catch most missing format strings.

Audit tip: Check for missing format strings for the following functions:

CFStringCreateWithFormat
CFStringCreateWithFormatAndArguments
CFStringAppendFormat
[NSString stringWithFormat:] and other NSString methods that take formatted strings as arguments:
- [NSString initWithFormat:]
- [NSString *WithFormat]
- [NSString stringByAppendingFormat]
- appendingFormat
- Wrong example:
  [x stringByAppendingFormat:[UtilityClass formatStuff:attacker.text]];
- Correct example:
  [x stringByAppendingFormat:@"%@", [UtilityClass formatStuff:attacker.text]];
[NSMutableString appendFormat]
[NSAlert alertWithMessageText]
[NSPredicate predicateWithFormat:]
[NSPredicate predicateWithFormat:arguments:]
[NSException raise:format:] and [NSException raise:format:arguments:]
NSRunAlertPanel and other Application Kit functions that create or return panels or sheets
[NSLog]

Security considerations for apps built with Swift

Keep the following in mind if you're developing iOS apps with Swift:

Swift uses Automatic Reference Counting (ARC) by default, which is very helpful.
If string interpolation is used, there is no risk of a format string attack.
An integer overflow causes a runtime error.
Buffer overflows generally cannot occur due to lack of pointers, except when UnsafePointer is used for C-compatibility.

Also, when handling sensitive memory, be aware that Swift won’t easily let you erase sensitive data, e.g. passwords. One way to do this is to use UnsafeMutablePointer or an UnsafeCollection (see Secure Memory for Swift Objects for more information).

Guide: Where should I store my data on iOS?

Where can I store my data?

Keychain Services
- Encrypted key/value store designed to hold:
  - Generic passwords
  - Internet passwords (password + protocol + server)
  - Certificates
  - Private Keys
  - Identities (certificate + private key)
- Max raw value size is ~16MB.
- Keychains may be shared (this is how SSO works on iOS) or private to the app.
  - Keychains can only be shared by apps from the same vendor.
  - Enterprise/Dogfood apps have a different vendor ID compared to Prod.

Your application has access to its own app-specific filesystem sandbox; please refer to Apple’s File System Programming Guide (specifically, the iOS sections) for more details.

Documents/
- User-created data that should be visible to the user
- Optionally visible to the user in iTunes
  - Subdirectories generally aren’t, special tools can still open them
- Backed up
  - User can disable backup for specific apps
  - App can disable paths by setting NSURLIsExcludedFromBackupKey
Library/Caches/
- Semi-persistent cached files
- Not visible to the user
- Not backed up
- May be deleted by the OS at any time if the app is not running
  - Managed automatically in response to storage pressure
Library/Application Support/
- Persistent files necessary to run the app
- Not visible to the user
- Backed up
  - User can disable backup for specific apps
  - App can disable paths by setting NSURLIsExcludedFromBackupKey
Library/Preferences/
- As /Application Support/
- By convention, only files created with NSUserDefaults
Library/*
- As /Application Support/
tmp/
- Non-persistent cached files
- Not visible to the user
- Not backed up
- Periodically deleted by the OS when the app is not running

Does the OS protect the keychain? How?

The keychain, on modern iOS devices (post-Touch ID) is secured using a hardware module. There are no known attacks that directly compromise the keychain via hardware or software; jailbroken devices are vulnerable to certain attacks.

Keychain backups (to iCloud) cannot be recovered without the user’s iCloud password. Keychain data is not included in local backups unless that backup is encrypted with a password.

Does the OS protect my files on disk? How?

Yes, the OS provides four levels of protection. Note that backups to iCloud are always encrypted and that backups in iTunes are optionally encrypted; unencrypted backups do not back up data marked in any of the protected classes below. The device’s filesystem is encrypted on modern iOS on the DMA path; these options add extra layers of security.

NSFileProtectionComplete - most secure
- Only readable if device is unlocked.
- File is closed when the device is locked.
- Suitable for most apps and data.
NSFileProtectionCompleteUnlessOpen
- File can only be opened when the device is unlocked.
- File is not closed when the device is locked.
- File is encrypted when the last open handle is closed.
- Suitable for data that is uploaded in the background, etc.
NSFileProtectionCompleteUntilFirstUserAuthentication (default)
- File is inaccessible until the device is unlocked once after boot.
- Suitable for background processes that should start ASAP after boot.
  - Geofence data
  - Bluetooth accessories (e.g. Android Wear)
- In general, all user data should be at least at this level.
NSFileProtectionNone - least secure
- No protection.
- Suitable for certain applications that must access data immediately on boot without any user interaction. This encryption/decryption is handled by the OS and the keychain transparently. The relevant decryption key is created from the keychain when appropriate and erased from memory when appropriate; see this guide for more details.

Where should I store my data?

Sensitive and persistent data - credentials, tokens, etc? Keychain.
Large sensitive and persistent files?
- Save it to the Library/* directory.
- Exclude it from backups.
  - Keychain backups have a higher level of security than filesystem backups.
- Set appropriate encryption options - as secure as possible.
Sensitive cache data?
- Save it to Library/Caches/*
- Set appropriate encryption options - as secure as possible.
Application configuration?
- NSUserDefaults? Library/Preferences/[Name].plist
- Other/custom format? Library/Application Support/*
- Set appropriate encryption options - as secure as possible.
Persistent content that should be backed up?
- User-generated and user-visible?
  - Documents/* directory.
  - Don’t use subdirectories if you want users to use iTunes file sharing.
  - NSFileProtectionCompleteUntilFirstUserAuthentication is probably the most appropriate option for encryption, if desired.
    - Note that malware on a trusted computer can access this directory if iTunes file sharing is enabled.
- Shouldn’t be visible to the user?
  - Library/Application Support/*
  - Set appropriate encryption options.

Best practices for storage

Store files securely

A stolen or lost iOS device can be potentially jailbroken or disassembled and the contents of the local file system can be read. Therefore iOS app developers need to make sure to encrypt sensitive information like credentials or other private information.

Keychain already allows you to prevent items from ever leaving the device or be included in backups.

In addition to that:

Items can be made to require user consent when accessed;
That consent can be set to Touch ID with the device password as fallback;
Items can be made inaccessible if passcode is removed.

The safest scenario would require flagging items as device-only, requiring Touch ID for access, and invalidated if passcode is ever removed.

Remember: you can also store any piece of text in Keychain, not just username and password credentials. Apple uses this to synchronize Wifi credentials between devices so that when you connect your laptop to a network, your phone will be able to as well a few seconds later when synchronization finishes, saving you from entering those long passwords on your phone. For more information on the details check out the Apple iOS Security white paper.

Audit tip: Check for stored data which is not using kSecAttrAccessibleWhenUnlocked or kSecAttrAccessibleAfterFirstUnlock. For example, if it is using kSecAttrAccessibleAlways, then the data is not sufficiently protected.

Audit tip: Check for files created with NSFileProtectionNone - they have no protection. Note that files created without explicit protection do not necessarily use NSFileProtectionNone. Make sure one of the following is used:
NSFileProtectionComplete NSFileProtectionCompleteUnlessOpen (key stays in memory while locked and file opened) NSFileProtectionCompleteUntilFirstUserAuthentication (key stays in memory when locked)

Create secure temporary files

Audit tip: Check that secure temporary files and directories are used - for example, URLForDirectory, NSTemporaryDirectory, FSFindFolder(kTemporaryFolderType). See also Create Temporary Files Correctly in the Apple Secure Coding Guide.

Avoid insecure destination files and APIs

Audit tip: Check for private information (PII) in NSLog/Alog, plist or local sqlite databases. It may not be encrypted. Logging is encrypted as of iOS 10.

Audit tip: Check that only appropriate user-specific non-sensitive information is written to iCloud storage. Use NSURLIsExcludedFromBackupKey to prevent backup of files to iCloud and iTunes.

Audit tip: For the Keychain, check that kSecAttrSynchronizable is false if the item is not intended for iCloud Keychain backup (it is false by default).

Audit tip: Check that NSUserDefaults does only contain settings and no personal information.

Testing for Devices with Jaibreak

Checking whether a device is jailbroken can be helpful to make certain in-app security decisions. Attackers can run tools like Cycript, GDB, or Snoop-it to perform runtime analysis and steal sensitive data from within your application. Jailbreak detection can prevent that.

Audit tip: Test that the app is not working on jaibroken devices.

Reference: https://github.com/felixgr/secure-ios-app-dev

Jason MillerFebruary 4, 2019

Latest Posts