パッカーの克服とOEPの特定

You understand obfuscation, can bypass anti-analysis checks, and grasp kernel-mode debugging. A common obstacle remains: packers that compress or encrypt the real code so static tools see only a stub.

A packer wraps the original program. At runtime a small unpacking stub decompresses or decrypts the real code into memory, then jumps to it.

• Smaller file size
• Hidden strings and imports
• Defeats naive static analysis

Signs of packing:

• High entropy sections (looks random)
• Few imports, odd section names like UPX0
• Tiny code region with a large memory allocation

Tools like Detect It Easy or PEiD flag known packers.

For well-known packers, a tool can reverse the process directly. UPX, for instance, has a built-in decompressor.

Custom or modified packers have no public unpacker. Then you let the stub do the work: run it under a debugger until the real code is in memory, then capture it.

This is generic, manual unpacking.

The OEP (Original Entry Point) is where the unpacked program's real execution begins. The stub jumps there after unpacking.

Finding the OEP is the key milestone: at that moment, the real code is fully unpacked in memory.

Stubs typically end with a far jump or push/ret into the unpacked region (the tail jump). Set a breakpoint there; when it fires, the next instruction is the OEP.

Another technique: set a hardware breakpoint on execute for the region the stub writes code into. Execution stops the instant the unpacked code runs.

ESP/stack-based tricks (the 'pushad/popad' method) also locate the tail.

At the OEP, dump the in-memory image to disk with a tool like Scylla or a debugger plugin.

The dump contains decrypted code and strings, but the import table is broken because it was resolved at runtime.

The final step is IAT reconstruction: tools like Scylla scan memory for the resolved imports and rebuild a valid Import Address Table, producing a clean, statically-analyzable executable.

Tough samples stack several packers. After dumping, your unpacked image may itself be packed again. Re-run detection on the dump.

Repeat the run-to-OEP-and-dump cycle until entropy drops and real strings and imports finally appear.

When manually unpacking, why is reaching the OEP the critical moment to dump the process?

You can now strip packers off protected binaries:

• Detect packing via entropy, sections, and imports
• Use known unpackers or run the stub to the OEP
• Find the tail jump, dump at the OEP, rebuild the IAT

The result is a clean image ready for full static analysis.