fsop-dev
Auth/lab dev: glibc FILE/FSOP exploitability research; libio structures, vtables, wide data, trigger paths, mitigation-aware modeling.
v2026.11.13-21cd5683 installs+3 in 7 days39 StarsMITsynced 11 hours ago
Install
No NotHoly CLI yet? Install it first.
$ curl -fsSL https://notholy.com/install.sh | sh$ notholy skills install aeondave/fsop-devWrites to ~/.notholy/skills after checking the signature and the SHA-256.
SignatureSigned, ed25519
Package31cd4603…2ddab010
31cd460383c4…436b2ddab0105 files · 15.0 KBScanLowno findings
Sourcegithub.com/AeonDave/malskill/tree/main/offensive-coding/fsop-dev@21cd568
On this page
Goal: turn FILE-structure corruption into the right libio dispatch path for that glibc era, trigger surface, and endgame — not into random _IO_FILE cargo cult.
Repository positioning: keep offensive-coding/heap-exploitation-dev as the allocator and overlap router, then switch here once success depends more on libio structure modeling, validated jump-table reuse, wide/codecvt/cookie-file dispatch, or trigger choice than on heap choreography itself. Engagement methodology and mitigation strategy live in offensive-techniques/binary-exploitation-technique; concrete FSOP recipes in offensive-ctf/pwn-ctf/references/heap-fsop.md.
When to activate
- Heap, UAF, overlap, arbitrary write, partial overwrite, or stream-adjacent corruption reaches
stdin,stdout,stderr, anotherFILE *, or a fakeFILEregion. - The exploit reaches a heap-allocated or dangling stream from
fopen/fdopen/customFILE *, and the real question becomes howfputs/fwrite/cleanup will consume it. - Need to decide whether the target is classic
_IO_list_all, vtable-misalignment,_wide_vtable,_codecvt, obstack, or leak-only FILE abuse. - Hooks are removed or unrealistic, and the real exploit question is whether FSOP beats return-address, callback, or data-only alternatives.
- Need to reason about
FILE,_IO_FILE_plus,_IO_wide_data,_IO_codecvt, pointer-guard-adjacent surfaces, or valid jump-table placement. - Need a hint-style recognition pass for stream corruption opportunities before building a full heap chain.
- The initial bug is not purely “heap-themed” anymore — for example, a relative libc write into a standard stream, an mmapped chunk reaching libc FILE data, or a direct stream-specific corruption bug.
If the problem is mostly allocator selection, heap shaping, tcache/largebin choreography, or House-family routing before the stream overlap exists, start with offensive-coding/heap-exploitation-dev and return here once FSOP is a real candidate.
First classify the FSOP problem
- Identify the stream target
stdoutfor output-driven paths likeputs,printf,fflushstderrfor low-noise or assert/abort-driven pathsstdinfor leak or arbitrary-read/write redirection- custom
FILE */fopenobject / socket-backed stream / heap fake FILE
- Identify the glibc era
- pre-2.24: no vtable validation
- 2.24+: vtable range validation
- 2.34+: hooks removed, FSOP becomes first-class endgame
- 2.35+ / modern:
_wide_vtable, codecvt, and angr-mapped paths matter more than fake raw heap vtables
- Decide the capability you actually need
- leak arbitrary memory
- arbitrary read/write through buffer redirection
- direct PC control / indirect call
setcontextpivot / ORW / SUID-safe chain- exit-time or assert-time execution
- Map the available corruption
- full stream overlap / arbitrary write into stream
- only a few stable fields
- only
vtable-adjacent control - only
_wide_dataor_codecvtreachability - only trigger-side influence with limited field control
- Choose the natural trigger surface
puts,printf,fprintf,fwritefflush,fclose,setbuf, sync/finish pathsexit,return from main, abort/assert,__malloc_assert- read-side paths like
fgets, underflow/seekoff, orientation switches
Era-first selection
| Era | Bias toward | Avoid or de-prioritize |
|---|---|---|
| glibc <= 2.23 | classic fake vtable, _IO_list_all, _IO_OVERFLOW, raw FILE-chain abuse |
post-hook-only thinking |
| glibc 2.24-2.33 | vtable-misalignment, validated jump-table reuse, _wide_vtable routes, leak-first stream abuse |
fake heap vtable pointers outside libc vtable section |
| glibc 2.34+ | House of Apple 2/3, _wide_vtable, codecvt, stderr/stdout overlap, setcontext/ORW endings |
__free_hook nostalgia |
| glibc 2.35+ and modern | angry-FSROP paths, _IO_wdoallocbuf, _IO_switch_to_wget_mode, codecvt, obstack, TLS/pointer-guard-adjacent pivots |
assuming only one Apple chain exists |
Technique families worth prioritizing
| Family | Good fit | Typical result | Reality check |
|---|---|---|---|
classic _IO_list_all FSOP |
pre-vtable-check or controlled valid vtable path | exit-time _IO_overflow dispatch |
foundational, but not the modern default |
| stdout/stderr overlap | arbitrary allocation/write over standard streams | trigger via normal output, fflush, or assert path |
still one of the best practical setups |
| buffer redirection | control _IO_read_* / _IO_write_* ranges |
arbitrary read/write or targeted leak | not all FSOP ends in RIP control |
| vtable misalignment | validated vtable section but wrong slot alignment | call a different libc libio function than intended | central post-2.24 concept |
| House of Apple 2 | _wide_data->_wide_vtable reachable |
direct call, system, or setcontext |
primary post-hook baseline |
| House of Apple 3 | _codecvt path easier than wide-data path |
indirect call through codecvt helper | higher setup cost, but powerful |
| obstack paths | stream or neighboring stream state controllable | call via _obstack_newchunk |
underused but real |
| leak-oriented FSOP | need TLS/libc/stack/tcb leak more than immediate RCE | stdout/stderr disclosure, pointer-guard recovery | often better than forcing shell-first |
| pointer-guard-adjacent FSOP | TLS/pointer-guard or encrypted callback surfaces in play | unlock later destructor/cookie-file hijack | usually a bridge, not the first stage |
| input-buffer redirection | can't forge in place (zeroing alloc) or redirect a RO stream pointer; an input stream is read each loop | corrupt only _IO_buf_base/read ptrs (keep vtable+_lock) → program's fgets/fread read()s a forged FILE into another stream → trigger on next print |
the practical escape from the in-place wall; CET/IBT-safe (dispatch stays in libc) |
Hint-mode recognition rules
Load references/hints-and-recognition.md when triaging quickly. The shortest recognition cues are:
- Triage the trigger first: scan imports/PLT — only
_exit(noexit/atexit) means_IO_list_allnever flushes; unusedstderrmeans no assert/error trigger. Don't forge for an unreachable trigger. - If you can overlap
stdoutorstderr, ask which natural call path touches it next before chasing ROP. - If the
vtablemust stay inside libc, ask which misaligned slot or alternate jump table gives the call you want. - If
_wide_datais reachable, ask can I force_IO_wdoallocbufor_IO_WOVERFLOW? - If
_codecvtis reachable, ask is this really Apple 3 / codecvt-in/out/length? - If a heap exploit already gives largebin/tcache positioning over a stream, bias toward FSOP before exotic leakless Houses.
- If seccomp or SUID makes
system("/bin/sh")weak, bias towardsetcontext, ORW, or a leak-first destructor chain. - If normal output mangles your crafted state, prefer
stderror assert-time dispatch. - In-place wall — if your only write primitive zeroes the chunk and the program does a FILE op on that stream before you can re-forge it (e.g. an alloc that
memsets thenprintfs), do not forge in place. Use cross-stream buffer redirection (corrupt an input stream's_IO_buf_base/read ptrs, keepvtable+_lock, let itsfgets/freadread()your forged FILE into another stream), pick a stream the corrupting op never touches, or use a non-zeroing/print-free/second-thread write.
Trigger-driven rules
puts/printf/fprintf: excellent when stdout is the target and you can satisfy byte-oriented path checks.fflush/ sync / finish: strong when you want a quieter dispatch point or an internal sync path.exit/ return frommain: best when_IO_list_alltraversal or cleanup logic is your dispatch engine.- abort /
__malloc_assert: strong when only failure-driven flushing is realistic; think Cat/Kiwi/Apple-style assert triggers. - read-side functions: use when the win is disclosure or when underflow/seekoff/orientation paths reach
_wide_vtableor codecvt.
Endgame selection
- Need simplest post-hook RCE: House of Apple 2 first.
- Need SUID-safe or seccomp-safe execution:
setcontextpivot, ORW, or structured ROP oversystem. - Need arbitrary disclosure or pointer-guard recovery: leak-oriented stdout/stderr FSOP first, execution second.
- Need valid-target, libc-internal dispatch with few writes: pick a misalignment or angry-FSROP path with the smallest stable field set.
- Have a cheaper application-owned callback overwrite already: do not force FSOP just because it is glamorous.
Reliability checklist
- Confirm the exact glibc version and whether your offsets match the target build.
- Separate overlap/corruption, dispatch path, and endgame.
- Keep
_lockvalid and writable unless the chosen path proves it will not be touched. - Track whether
_modemust stay negative, become positive, or remain zero for the intended path. - Re-check whether read/write helpers will clobber
_IO_read_*,_IO_write_*,_wide_data, or adjacent bytes before dispatch. - Verify whether the chosen path expects byte-oriented, wide-oriented, or codecvt-backed state.
- Prefer the stream whose normal use interferes least with your crafted layout.
- If the chain rewrites a libc GOT/relocation target or relies on a low-byte pointer patch, load
pwn-ctfRELRO/ASLR relocation guidance and prove object-specific RELRO, page permissions, and ASLR invariance first.
Anti-patterns
- Treating every FILE corruption as House of Apple 2.
- Overwriting a vtable pointer to raw heap memory on glibc with validation enabled.
- Forgetting that many FSOP wins are leak/read/write primitives, not immediate PC control.
- Recommending
system("/bin/sh")on SUID or seccomp targets by reflex. - Ignoring normal stream activity that mutates
_IO_write_ptr,_IO_buf_base,_mode, or_flagsbefore the trigger fires. - Forging a FILE for an unreachable trigger (e.g.
_IO_list_allwhen the binary only_exits) — verify the trigger in the imports/PLT first. - Repeatedly attempting an in-place forge when the write primitive zeroes the stream and the program immediately uses it — that ordering can never complete; switch to buffer redirection or a print-free write.
Resources
- references/libio-model-and-eras.md —
FILE,_IO_FILE_plus,_IO_wide_data,_IO_codecvt, vtable validation, and era map. - references/canonical-fsop-families.md — classic
_IO_list_all, Apple 2/3, Pig, Cat, Kiwi, Emma, and where angry-FSROP paths fit. - references/triggers-and-call-paths.md —
puts,printf,fflush,exit, assert/abort, underflow, seekoff, and why some paths are cleaner than others. - references/hints-and-recognition.md — fast hint-mode triage, constraint heuristics, target-selection rules, and endgame cues.
Load references only after the target stream, glibc era, available corruption, and intended trigger are clear.
Files in this version
references/430.4 KB
- canonical-fsop-families.md7.6 KB
- hints-and-recognition.md6.9 KB
- libio-model-and-eras.md5.6 KB
- triggers-and-call-paths.md10.2 KB
- SKILL.md11.5 KB
Permissions
| Permission | Declared | Found by the scan |
|---|---|---|
| Reads files | Yes | Yes |
| Writes files | No | No |
| Runs shell commands | No | No |
| Uses the network | No | No |
| Uses connectors | No | No |
| Drives a browser | No | No |
Scan findings
Reviewed on Nov 13, 2026
The scan found nothing to flag in this version.
A clean scan lowers the risk but does not prove a package is safe. Read what you install.
| Version | Published | Commit | SHA-256 | Signature | Permission change |
|---|---|---|---|---|---|
| v2026.11.13-21cd568 | Nov 13, 2026 | 21cd568 | 31cd4603 | Signed | – |
More from AeonDave
All from this publisherMore in Reverse engineering and fuzzing
See the categoryAsk anything your scope allows.
Create an account, add credits and ask the first question in the chat or the terminal.
- 1/3
Create the account
Individual or company. You can switch to a company later without losing balance or keys.
IndividualCompany - 2/3
Add credits
By card. Pay per token, no subscription.
$5.00minimum deposit - 3/3
Ask
In the chat, or install the CLI and run notholy login in your terminal.
$ notholy login