macOS Universal binaries & Mach-O Format

Learn AWS hacking from zero to hero with htARTE (HackTricks AWS Red Team Expert)!

Other ways to support HackTricks:

Basic Information

Mac OS binaries usually are compiled as universal binaries. A universal binary can support multiple architectures in the same file.

These binaries follows the Mach-O structure which is basically compased of:

  • Header

  • Load Commands

  • Data

Fat Header

Search for the file with: mdfind fat.h | grep -i mach-o | grep -E "fat.h$"

#define FAT_MAGIC	0xcafebabe
#define FAT_CIGAM	0xbebafeca	/* NXSwapLong(FAT_MAGIC) */

struct fat_header {
	uint32_t	magic;		/* FAT_MAGIC or FAT_MAGIC_64 */
	uint32_t	nfat_arch;	/* number of structs that follow */

struct fat_arch {
	cpu_type_t	cputype;	/* cpu specifier (int) */
	cpu_subtype_t	cpusubtype;	/* machine specifier (int) */
	uint32_t	offset;		/* file offset to this object file */
	uint32_t	size;		/* size of this object file */
	uint32_t	align;		/* alignment as a power of 2 */

The header has the magic bytes followed by the number of archs the file contains (nfat_arch) and each arch will have a fat_arch struct.

Check it with:

% file /bin/ls
/bin/ls: Mach-O universal binary with 2 architectures: [x86_64:Mach-O 64-bit executable x86_64] [arm64e:Mach-O 64-bit executable arm64e]
/bin/ls (for architecture x86_64):	Mach-O 64-bit executable x86_64
/bin/ls (for architecture arm64e):	Mach-O 64-bit executable arm64e

% otool -f -v /bin/ls
Fat headers
fat_magic FAT_MAGIC
nfat_arch 2
architecture x86_64
    cputype CPU_TYPE_X86_64
    cpusubtype CPU_SUBTYPE_X86_64_ALL
    capabilities 0x0
    offset 16384
    size 72896
    align 2^14 (16384)
architecture arm64e
    cputype CPU_TYPE_ARM64
    cpusubtype CPU_SUBTYPE_ARM64E
    offset 98304
    size 88816
    align 2^14 (16384)

or using the Mach-O View tool:

As you may be thinking usually a universal binary compiled for 2 architectures doubles the size of one compiled for just 1 arch.

Mach-O Header

The header contains basic information about the file, such as magic bytes to identify it as a Mach-O file and information about the target architecture. You can find it in: mdfind loader.h | grep -i mach-o | grep -E "loader.h$"

#define	MH_MAGIC	0xfeedface	/* the mach magic number */
#define MH_CIGAM	0xcefaedfe	/* NXSwapInt(MH_MAGIC) */
struct mach_header {
	uint32_t	magic;		/* mach magic number identifier */
	cpu_type_t	cputype;	/* cpu specifier (e.g. I386) */
	cpu_subtype_t	cpusubtype;	/* machine specifier */
	uint32_t	filetype;	/* type of file (usage and alignment for the file) */
	uint32_t	ncmds;		/* number of load commands */
	uint32_t	sizeofcmds;	/* the size of all the load commands */
	uint32_t	flags;		/* flags */

#define MH_MAGIC_64 0xfeedfacf /* the 64-bit mach magic number */
#define MH_CIGAM_64 0xcffaedfe /* NXSwapInt(MH_MAGIC_64) */
struct mach_header_64 {
	uint32_t	magic;		/* mach magic number identifier */
	int32_t		cputype;	/* cpu specifier */
	int32_t		cpusubtype;	/* machine specifier */
	uint32_t	filetype;	/* type of file */
	uint32_t	ncmds;		/* number of load commands */
	uint32_t	sizeofcmds;	/* the size of all the load commands */
	uint32_t	flags;		/* flags */
	uint32_t	reserved;	/* reserved */

Mach-O File Types

There are different file types, you can find them defined in the source code for example here. The most important ones are:

  • MH_OBJECT: Relocatable object file (intermediate products of compilation, not executables yet).

  • MH_EXECUTE: Executable files.

  • MH_FVMLIB: Fixed VM library file.

  • MH_CORE: Code Dumps

  • MH_PRELOAD: Preloaded executable file (no longer supported in XNU)

  • MH_DYLIB: Dynamic Libraries

  • MH_DYLINKER: Dynamic Linker

  • MH_BUNDLE: "Plugin files". Generated using -bundle in gcc and explicitly loaded by NSBundle or dlopen.

  • MH_DYSM: Companion .dSym file (file with symbols for debugging).

  • MH_KEXT_BUNDLE: Kernel Extensions.

# Checking the mac header of a binary
otool -arch arm64e -hv /bin/ls
Mach header
      magic  cputype cpusubtype  caps    filetype ncmds sizeofcmds      flags
MH_MAGIC_64    ARM64          E USR00     EXECUTE    19       1728   NOUNDEFS DYLDLINK TWOLEVEL PIE

Or using Mach-O View:

Mach-O Flags

The source code also defines several flags useful for loading libraries:

  • MH_NOUNDEFS: No undefined references (fully linked)

  • MH_DYLDLINK: Dyld linking

  • MH_PREBOUND: Dynamic references prebound.

  • MH_SPLIT_SEGS: File splits r/o and r/w segments.

  • MH_WEAK_DEFINES: Binary has weak defined symbols

  • MH_BINDS_TO_WEAK: Binary uses weak symbols

  • MH_ALLOW_STACK_EXECUTION: Make the stack executable


  • MH_PIE: Position Independent Executable

  • MH_HAS_TLV_DESCRIPTORS: There is a section with thread local variables

  • MH_NO_HEAP_EXECUTION: No execution for heap/data pages

  • MH_HAS_OBJC: Binary has oBject-C sections

  • MH_SIM_SUPPORT: Simulator support

  • MH_DYLIB_IN_CACHE: Used on dylibs/frameworks in shared library cache.

Mach-O Load commands

The file's layout in memory is specified here, detailing the symbol table's location, the context of the main thread at execution start, and the required shared libraries. Instructions are provided to the dynamic loader (dyld) on the binary's loading process into memory.

The uses the load_command structure, defined in the mentioned loader.h:

struct load_command {
        uint32_t cmd;           /* type of load command */
        uint32_t cmdsize;       /* total size of command in bytes */

There are about 50 different types of load commands that the system handles differently. The most common ones are: LC_SEGMENT_64, LC_LOAD_DYLINKER, LC_MAIN, LC_LOAD_DYLIB, and LC_CODE_SIGNATURE.


Basically, this type of Load Command define how to load the __TEXT (executable code) and __DATA (data for the process) segments according to the offsets indicated in the Data section when the binary is executed.

These commands define segments that are mapped into the virtual memory space of a process when it is executed.

There are different types of segments, such as the __TEXT segment, which holds the executable code of a program, and the __DATA segment, which contains data used by the process. These segments are located in the data section of the Mach-O file.

Each segment can be further divided into multiple sections. The load command structure contains information about these sections within the respective segment.

In the header first you find the segment header:

struct segment_command_64 { /* for 64-bit architectures */
	uint32_t	cmd;		/* LC_SEGMENT_64 */
	uint32_t	cmdsize;	/* includes sizeof section_64 structs */
	char		segname[16];	/* segment name */
	uint64_t	vmaddr;		/* memory address of this segment */
	uint64_t	vmsize;		/* memory size of this segment */
	uint64_t	fileoff;	/* file offset of this segment */
	uint64_t	filesize;	/* amount to map from the file */
	int32_t		maxprot;	/* maximum VM protection */
	int32_t		initprot;	/* initial VM protection */
	uint32_t	nsects;		/* number of sections in segment */
	uint32_t	flags;		/* flags */

Example of segment header:

This header defines the number of sections whose headers appear after it:

struct section_64 { /* for 64-bit architectures */
	char		sectname[16];	/* name of this section */
	char		segname[16];	/* segment this section goes in */
	uint64_t	addr;		/* memory address of this section */
	uint64_t	size;		/* size in bytes of this section */
	uint32_t	offset;		/* file offset of this section */
	uint32_t	align;		/* section alignment (power of 2) */
	uint32_t	reloff;		/* file offset of relocation entries */
	uint32_t	nreloc;		/* number of relocation entries */
	uint32_t	flags;		/* flags (section type and attributes)*/
	uint32_t	reserved1;	/* reserved (for offset or index) */
	uint32_t	reserved2;	/* reserved (for count or sizeof) */
	uint32_t	reserved3;	/* reserved */

Example of section header:

If you add the section offset (0x37DC) + the offset where the arch starts, in this case 0x18000 --> 0x37DC + 0x18000 = 0x1B7DC

It's also possible to get headers information from the command line with:

otool -lv /bin/ls

Common segments loaded by this cmd:

  • __PAGEZERO: It instructs the kernel to map the address zero so it cannot be read from, written to, or executed. The maxprot and minprot variables in the structure are set to zero to indicate there are no read-write-execute rights on this page.

    • This allocation is important to mitigate NULL pointer dereference vulnerabilities. This is because XNU enforces a hard page zero that ensures the first page (only the first) of memory is innaccesible (except in i386). A binary could fulfil this requirements by crafting a small __PAGEZERO (using the -pagezero_size) to cover the first 4k and having the rest of 32bit memory accessible in both user and kernel mode.

  • __TEXT: Contains executable code with read and execute permissions (no writable). Common sections of this segment:

    • __text: Compiled binary code

    • __const: Constant data (read only)

    • __[c/u/os_log]string: C, Unicode or os logs string constants

    • __stubs and __stubs_helper: Involved during the dynamic library loading process

    • __unwind_info: Stack unwind data.

    • Note that all this content is signed but also marked as executable (creating more options for exploitation of sections that doesn't necessarily need this privilege, like string dedicated sections).

  • __DATA: Contains data that is readable and writable (no executable).

    • __got: Global Offset Table

    • __nl_symbol_ptr: Non lazy (bind at load) symbol pointer

    • __la_symbol_ptr: Lazy (bind on use) symbol pointer

    • __const: Should be read-only data (not really)

    • __cfstring: CoreFoundation strings

    • __data: Global variables (that have been initialized)

    • __bss: Static variables (that have not been initialized)

    • __objc_* (__objc_classlist, __objc_protolist, etc): Information used by the Objective-C runtime

  • __DATA_CONST: __DATA.__const is not guaranteed to be constant (write permissions), nor are other pointers and the GOT. This section makes __const, some initializers and the GOT table (once resolved) read only using mprotect.

  • __LINKEDIT: Contains information for the linker (dyld) such as, symbol, string, and relocation table entries. It' a generic container for contents that are neither in __TEXT or __DATA and its content is decribed in other load commands.

    • dyld information: Rebase, Non-lazy/lazy/weak binding opcodes and export info

    • Functions starts: Table of start addresses of functions

    • Data In Code: Data islands in __text

    • SYmbol Table: Symbols in binary

    • Indirect Symbol Table: Pointer/stub symbols

    • String Table

    • Code Signature

  • __OBJC: Contains information used by the Objective-C runtime. Though this information might also be found in the __DATA segment, within various in __objc_* sections.

  • __RESTRICT: A segment without content with a single section called __restrict (also empty) that ensures that when running the binary, it will ignore DYLD environmental variables.

As it was possible to see in the code, segments also support flags (although they aren't used very much):

  • SG_HIGHVM: Core only (not used)

  • SG_FVMLIB: Not used

  • SG_NORELOC: Segment has no relocation

  • SG_PROTECTED_VERSION_1: Encryption. Used for example by Finder to encrypt text __TEXT segment.


LC_MAIN contains the entrypoint in the entryoff attribute. At load time, dyld simply adds this value to the (in-memory) base of the binary, then jumps to this instruction to start execution of the binary’s code.

LC_UNIXTHREAD contains the values the register must have when starting the main thread. This was already deprecated but dyld still uses it. It's possible to see the vlaues of the registers set by this with:

otool -l /usr/lib/dyld
Load command 13
        cmd LC_UNIXTHREAD
    cmdsize 288
     flavor ARM_THREAD_STATE64
	    x0  0x0000000000000000 x1  0x0000000000000000 x2  0x0000000000000000
	    x3  0x0000000000000000 x4  0x0000000000000000 x5  0x0000000000000000
	    x6  0x0000000000000000 x7  0x0000000000000000 x8  0x0000000000000000
	    x9  0x0000000000000000 x10 0x0000000000000000 x11 0x0000000000000000
	    x12 0x0000000000000000 x13 0x0000000000000000 x14 0x0000000000000000
	    x15 0x0000000000000000 x16 0x0000000000000000 x17 0x0000000000000000
	    x18 0x0000000000000000 x19 0x0000000000000000 x20 0x0000000000000000
	    x21 0x0000000000000000 x22 0x0000000000000000 x23 0x0000000000000000
	    x24 0x0000000000000000 x25 0x0000000000000000 x26 0x0000000000000000
	    x27 0x0000000000000000 x28 0x0000000000000000  fp 0x0000000000000000
	     lr 0x0000000000000000 sp  0x0000000000000000  pc 0x0000000000004b70
	   cpsr 0x00000000



Contains information about the code signature of the Macho-O file. It only contains an offset that points to the signature blob. This is typically at the very end of the file. However, you can find some information about this section in this blog post and this gists.


Support for binary encryption. However, of course, if an attacker manages to compromise the process, he will be able to dump the memory unencrypted.


Contains the path to the dynamic linker executable that maps shared libraries into the process address space. The value is always set to /usr/lib/dyld. It’s important to note that in macOS, dylib mapping happens in user mode, not in kernel mode.


Obsolete but when configured to geenrate dumps on panic, a Mach-O core dump is created and the kernel version is set in the LC_IDENT command.


Random UUID. It's useful for anything directly but XNU caches it with the rest of the process info. It can be used in crash reports.


Allows to indicate environment variables to the dyld beforenthe process is executed. This can be vary dangerous as it can allow to execute arbitrary code inside the process so this load command is only used in dyld build with #define SUPPORT_LC_DYLD_ENVIRONMENT and further restricts processing only to variables of the form DYLD_..._PATH specifying load paths.


This load command describes a dynamic library dependency which instructs the loader (dyld) to load and link said library. There is a LC_LOAD_DYLIB load command for each library that the Mach-O binary requires.

  • This load command is a structure of type dylib_command (which contains a struct dylib, describing the actual dependent dynamic library):

struct dylib_command {
        uint32_t        cmd;            /* LC_LOAD_{,WEAK_}DYLIB */
        uint32_t        cmdsize;        /* includes pathname string */
        struct dylib    dylib;          /* the library identification */ 

struct dylib {
    union lc_str  name;                 /* library's path name */
    uint32_t timestamp;                 /* library's build time stamp */
    uint32_t current_version;           /* library's current version number */
    uint32_t compatibility_version;     /* library's compatibility vers number*/

You could also get this info from the cli with:

otool -L /bin/ls
	/usr/lib/libutil.dylib (compatibility version 1.0.0, current version 1.0.0)
	/usr/lib/libncurses.5.4.dylib (compatibility version 5.4.0, current version 5.4.0)
	/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 1319.0.0)

Some potential malware related libraries are:

  • DiskArbitration: Monitoring USB drives

  • AVFoundation: Capture audio and video

  • CoreWLAN: Wifi scans.

A Mach-O binary can contain one or more constructors, that will be executed before the address specified in LC_MAIN. The offsets of any constructors are held in the __mod_init_func section of the __DATA_CONST segment.

Mach-O Data

At the core of the file lies the data region, which is composed of several segments as defined in the load-commands region. A variety of data sections can be housed within each segment, with each section holding code or data specific to a type.

The data is basically the part containing all the information that is loaded by the load commands LC_SEGMENTS_64

This includes:

  • Function table: Which holds information about the program functions.

  • Symbol table: Which contains information about the external function used by the binary

  • It could also contain internal function, variable names as well and more.

To check it you could use the Mach-O View tool:

Or from the cli:

size -m /bin/ls

Objetive-C Common Sections

In __TEXT segment (r-x):

  • __objc_classname: Class names (strings)

  • __objc_methname: Method names (strings)

  • __objc_methtype: Method types (strings)

In __DATA segment (rw-):

  • __objc_classlist: Pointers to all Objetive-C classes

  • __objc_nlclslist: Pointers to Non-Lazy Objective-C classes

  • __objc_catlist: Pointer to Categories

  • __objc_nlcatlist: Pointer to Non-Lazy Categories

  • __objc_protolist: Protocols list

  • __objc_const: Constant data

  • __objc_imageinfo, __objc_selrefs, objc__protorefs...


  • _swift_typeref, _swift3_capture, _swift3_assocty, _swift3_types, _swift3_proto, _swift3_fieldmd, _swift3_builtin, _swift3_reflstr

Learn AWS hacking from zero to hero with htARTE (HackTricks AWS Red Team Expert)!

Other ways to support HackTricks:

Last updated