FLIRTシグネチャとライブラリ関数の識別

You can script disassemblers, automate structure recovery, and patch binaries. But statically-linked programs bundle thousands of library functions (libc, the C++ STL, runtime).

Wading through them by hand wastes enormous time.

When a binary is statically linked, library code is copied directly into the executable. There are no import names; printf just looks like another anonymous function.

Identifying these frees you to focus on the author's own code.

FLIRT (Fast Library Identification and Recognition Technology) is IDA's system for matching byte patterns of known library functions and auto-naming them.

Ghidra has an equivalent via Function ID databases.

A signature records a function's opcode bytes, masking out parts that vary (like relocated addresses).

The tool scans the binary; when bytes match a signature, it applies the known name and prototype.

IDA ships .sig files for common runtimes. You apply them from File, Load file, FLIRT signature file, then IDA renames matched functions.

Suddenly hundreds of sub_xxxx become recognizable like strcpy and malloc.

For uncommon or custom static libraries, generate signatures with IDA's FLAIR tools: parse the .a archive into a pattern file, then compile it to a .sig.

Ghidra's Function ID plugin hashes function bodies and stores them in a database. Importing a database for a known runtime auto-labels matches in your target.

You can build databases from libraries you have analyzed before.

Once libraries are named, your scripts can skip them. Iterate functions and ignore any tagged as library code, analyzing only user functions.

Signatures depend on the exact compiler and version. A different optimization level can prevent a match, and short functions may match the wrong library.

Always sanity-check auto-named functions before trusting them.

Combine signatures with string and xref analysis. A function FLIRT names printf should have format-string xrefs nearby; if not, the match may be wrong.

Cross-validation builds confidence.

Identifying a library function also imports its prototype. Once memcpy(dst, src, n) is recognized, the decompiler labels its three arguments correctly.

This propagates type information into callers, sharply improving pseudocode readability.

You can now cut through library clutter:

• Static linking hides libraries as anonymous functions
• FLIRT (IDA) and Function ID (Ghidra) auto-name them by pattern
• Build custom signatures with FLAIR for uncommon libs
• Script to skip library code, but verify matches