Manually unpacking a packed binary

Manually Unpacking a Packed Binary

When to Use

• A sample is runtime-packed by a custom/unknown packer that UPX and generic unpackers cannot handle.
• You need to locate the original entry point (OEP) and dump the unpacked image for static analysis.

Do not use generic auto-unpackers here — this is the manual path for cases where they fail. Run the sample only inside an isolated, instrumented debugger VM.

Prerequisites

• A debugger (x64dbg) on an isolated VM and the packed sample.

Safety & Handling

• The sample executes during unpacking; do this only in a disposable, isolated VM with snapshots.

Workflow

Step 1: Characterize the packer statically

python scripts/analyst.py inspect packed.exe

Reports section entropy, suspicious section names (UPX, .vmp, random), small import tables, and a write-then-execute section (classic unpacker stub).

Step 2: Find candidate OEP heuristics

Set a memory breakpoint on the unpacked code section's execution; the packer's tail jump (a far jmp/ret into a different section) typically transfers control to the OEP.

Step 3: Dump at OEP

When stopped at OEP, dump the process image and proceed to import reconstruction (see the dump-and-rebuild skill).

Step 4: Verify the dump

Confirm the dumped image has a sane entry point, restored imports, and readable code.

Validation

• The unpacking stub is identified by a write→execute section transition.
• The tail jump lands in a different section than the stub (OEP candidate).
• The dumped image disassembles cleanly with recognizable runtime startup code.

Pitfalls

• Stopping at a stage-2 stub rather than the true OEP in multi-layer packers.
• Hardware-breakpoint detection by the packer — vary the breakpoint strategy.
• Dumping before the import table is fully resolved.

References

• See references/api-reference.md for the inspector.
• ATT&CK T1027.002 and the PE format spec (linked in frontmatter).