tsk-arasu/executorch-mfv-poc-stride-integer-overflow

1# Signed Integer Overflow in ExecuTorch Tensor Stride Computation via Unchecked Size Product (.pte)2 3**Target:** ExecuTorch 1.3.1 (.pte, huntr Model File Vulnerability program)4**Severity:** Medium (integer overflow with plausible but unproven downstream memory-corruption chain; DoS confirmed)5**CWE:** CWE-190 (Integer Overflow or Wraparound)6**Component:** `runtime/core/exec_aten/util/dim_order_util.h` (overflow site) / `runtime/executor/tensor_parser_portable.cpp` (caller)7**Authentication Required:** No — only requires a victim application to load an attacker-supplied `.pte` file via the standard, always-used `Program::load_method()` API.8 9## Summary10 11ExecuTorch computes a tensor's memory strides from its `sizes` and `dim_order` fields during ordinary `.pte` tensor deserialization. The stride computation multiplies a running product of dimension sizes with no overflow check. A `.pte` file with a `Tensor` value whose `sizes` contains a large-but-individually-valid dimension (passing the existing "no negative sizes" check) causes this multiplication to overflow a 32-bit signed integer — Undefined Behavior in C++, caught deterministically by UBSan. This is the first finding across a broader security review of this codebase reached through the core `Method::init()` deserialization path (not a metadata-only accessor), and the first integer-overflow (rather than null-pointer-dereference) bug class identified.12 13huntr's own Model File Vulnerability program explicitly lists **"integer overflows"** as an example of "vulnerabilities in model file parsing leading to memory corruption" — this finding is a direct, verbatim match for that named category.14 15## Vulnerability Details16 17`runtime/executor/tensor_parser_portable.cpp`'s `parseTensor()` validates that no individual size is negative:18 19```cpp20for (flatbuffers::uoffset_t i = 0; i < dim; i++) {21  ET_CHECK_OR_RETURN_ERROR(22      sizes[i] >= 0,23      InvalidProgram,24      "Negative size[%zu] %" PRId32,25      static_cast<size_t>(i),26      sizes[i]);27}28```29 30This check does **not** bound the *product* of sizes against the 32-bit signed integer type used for strides. The tensor's strides are subsequently computed via `dim_order_to_stride()` → `dim_order_to_stride_nocheck()` in `runtime/core/exec_aten/util/dim_order_util.h`:31 32```cpp33template <typename SizesType, typename DimOrderType, typename StridesType>34inline void dim_order_to_stride_nocheck(35    const SizesType* sizes,36    const DimOrderType* dim_order,37    const size_t dims,38    StridesType* strides) {39  if (dims == 0) return;40  strides[dim_order[dims - 1]] = 1;41  for (int32_t i = dims - 2; i >= 0; --i) {42    if (sizes[dim_order[i + 1]] == 0) {43      strides[dim_order[i]] = strides[dim_order[i + 1]];44    } else {45      strides[dim_order[i]] =46          strides[dim_order[i + 1]] * sizes[dim_order[i + 1]];   // <-- unchecked multiplication, line 14747    }48  }49}50```51 52`StridesType`/`SizesType` default to `int32_t`. This multiplication has **no** overflow check, unlike the analogous computation in `runtime/core/tensor_layout.cpp`'s `calculate_nbytes()` (used by the parallel `.ptd` `TensorLayout` code path), which explicitly calls `c10::mul_overflows()` before trusting the product. The `.pte` `Tensor` stride-computation path has no equivalent protection.53 54## Steps to Reproduce55 56### Environment57Linux x86-64, ExecuTorch 1.3.1 pristine source, clang-16, CMake, Ninja. No authentication, no host access.58 59### 1. Build ExecuTorch with sanitizers60 61Same build as REPORT-01 Step 1.62 63### 2. Build the load_method-level harness (`poc/harness_load_method_fuzzer.cpp`, included in this report)64 65This is a new harness (not used in prior reports) that goes one layer deeper than metadata-only accessors — it calls `Program::load_method()`, exercising `Method::init()` → `parse_values()` → `parseTensor()`, the CORE tensor-deserialization path used every time a `.pte` is loaded:66 67```bash68export ET_PARENT=/path/to/parent-of-executorch69C10_INC="$ET_SRC/runtime/core/portable_type/c10"70INCLUDES="-I$ET_PARENT -I$ET_BUILD -I$ET_BUILD/schema/include -I$ET_BUILD/extension/flat_tensor/include -I$ET_BUILD/third-party/flatc_ep/include -I$C10_INC"71 72clang++-16 -std=c++17 -fsanitize=fuzzer,address,undefined -fno-omit-frame-pointer -fno-sanitize-recover=all \73  $INCLUDES -DFLATBUFFERS_MAX_ALIGNMENT=1024 -DC10_USING_CUSTOM_GENERATED_MACROS \74  -c poc/harness_load_method_fuzzer.cpp -o harness.o75 76clang++-16 -fsanitize=fuzzer,address,undefined -o poc_harness harness.o \77  "$ET_BUILD/extension/data_loader/libextension_data_loader.a" \78  "$ET_BUILD/libexecutorch_core.a"79```80 81### 3. PoC file82 83`poc/poc_stride_int_overflow.pte` (24,054 bytes, **included in this report — sha256 `0767af28fc00e63e8ea665beaf5139f7d84332abc435bd8e0de3ebf8d0936e98`**) is a `.pte` file found via coverage-guided fuzzing (within ~10,000 executions) containing a `Tensor` value whose `sizes` array produces a stride-computation overflow. The libFuzzer crash minimizer was unable to reduce this file below its original size while preserving the crash, so the original fuzzer-found file is the canonical reproducer.84 85### 4. Trigger the crash86 87```bash88export ASAN_OPTIONS="abort_on_error=1:symbolize=0"89export UBSAN_OPTIONS="halt_on_error=1:print_stacktrace=0"90./poc_harness -timeout=5 -runs=0 poc/poc_stride_int_overflow.pte91```92 93### Expected result (secure behavior)94`Program::load_method()` should return a clean `Error::InvalidProgram` when the size product would overflow the stride computation, matching the overflow-checked pattern already used in `calculate_nbytes()` for the parallel `.ptd` code path.95 96### Actual result — verified against the pristine, unmodified ExecuTorch 1.3.1 source97 98```99Running: poc/poc_stride_int_overflow.pte100runtime/core/exec_aten/util/dim_order_util.h:147:37: runtime error: signed integer overflow: 8 * 2113929312 cannot be represented in type 'int'101SUMMARY: UndefinedBehaviorSanitizer: undefined-behavior runtime/core/exec_aten/util/dim_order_util.h:147:37 in102==<pid>== ERROR: libFuzzer: deadly signal103    #0 ... (abort machinery)104    ...105    dim_order_to_stride_nocheck (dim_order_util.h:147)106    dim_order_to_stride (dim_order_util.h:168)107    deserialization::parseTensor (tensor_parser_portable.cpp:144)108```109 110Full symbolized call chain (via `addr2line` against the built binary) confirms:111`dim_order_to_stride_nocheck` (dim_order_util.h:147) ← `dim_order_to_stride` (dim_order_util.h:168) ← `deserialization::parseTensor` (tensor_parser_portable.cpp:144).112 113**Reproduced 3/3 identical runs against the pristine source** (re-verified live for this report):114```115run 1: signed integer overflow: 8 * 2113929312 cannot be represented in type 'int'116run 2: (identical)117run 3: (identical)118```119 120## Chain Escalation Attempt (Reported Honestly)121 122Per standard practice of testing whether a confirmed bug can be escalated to a stronger primitive: a second sanitized build was made with **UBSan disabled** (ASan only), so the signed-integer-overflow multiplication would wrap silently instead of aborting — this lets us observe what actually happens to the corrupted stride value downstream, rather than only knowing it aborts under a sanitizer.123 124**Result: the identical PoC ran cleanly with zero ASan errors** when UBSan wasn't present to trap the overflow. This means:125 126- The corrupted/wrapped stride value **is** computed and stored in the tensor's `strides` array.127- **Nothing within `Program::load_method()`'s own code path** (`parse_values`/`parseTensor`/`dim_order_to_stride`) subsequently dereferences or uses that stride to compute a memory address.128- The stride is only actually *used* for address computation when an **operator executes** and indexes into the tensor's data using it — a step that happens during `Method::execute()`, not `load_method()`.129 130**This report does not claim a proven memory-corruption chain.** Proving one would require a substantially larger harness — kernel registration via `register_kernels()`, a full `Method::execute()` invocation, and a `.pte` whose instruction chain actually invokes an operator against the malformed tensor — none of which was built in this investigation. The finding is reported as a **confirmed integer-overflow DoS** (deterministic crash under sanitizers; silent, unobserved UB in production builds) with a **plausible but unproven** downstream out-of-bounds access risk during subsequent operator execution.131 132## Impact133 134**Who is affected:** Any application calling `Program::load_method()` — the standard, always-used method-loading API — on a `.pte` containing a `Tensor` value with a crafted `sizes` array. This is the core deserialization path, not a rarely-exercised accessor.135 136**What the attacker can do (proven):** Cause a reliable, deterministic crash (SIGABRT under UBSan) purely by supplying a malformed model file.137 138**What the attacker might additionally be able to do (unproven, flagged honestly):** If the corrupted/wrapped stride value later feeds into address computation during operator execution, it could plausibly cause an out-of-bounds memory read or write — this was investigated but not demonstrated in this session; see Chain Escalation Attempt above.139 140**What's at risk:** Availability, confirmed. Integrity/memory-safety during operator execution, plausible but not demonstrated.141 142**Why Medium, not Critical:** Matches huntr's explicitly-named "integer overflow" example under the memory-corruption category, but only the DoS consequence was proven — claiming a full memory-corruption chain without a working PoC would be overclaiming.143 144## Suggested Remediation145 146```cpp147template <typename SizesType, typename DimOrderType, typename StridesType>148inline Error dim_order_to_stride_nocheck_safe(149    const SizesType* sizes,150    const DimOrderType* dim_order,151    const size_t dims,152    StridesType* strides) {153  if (dims == 0) return Error::Ok;154  strides[dim_order[dims - 1]] = 1;155  for (int32_t i = dims - 2; i >= 0; --i) {156    if (sizes[dim_order[i + 1]] == 0) {157      strides[dim_order[i]] = strides[dim_order[i + 1]];158    } else {159      StridesType product;160      if (c10::mul_overflows(strides[dim_order[i + 1]], sizes[dim_order[i + 1]], &product)) {161        return Error::InvalidArgument;162      }163      strides[dim_order[i]] = product;164    }165  }166  return Error::Ok;167}168```169 170This mirrors the overflow-checked pattern already correctly implemented in `runtime/core/tensor_layout.cpp`'s `calculate_nbytes()` via `c10::mul_overflows()`.171 172A regression test should build a `.pte` `Tensor` value whose `sizes` product overflows `INT32_MAX` while each individual size passes the existing non-negative check, asserting `parseTensor()` returns a clean `Error` rather than triggering UB.173 174## Files Included in This Report175 176- `poc/poc_stride_int_overflow.pte` — the 24,054-byte PoC file (sha256 `0767af28fc00e63e8ea665beaf5139f7d84332abc435bd8e0de3ebf8d0936e98`)177- `poc/harness_load_method_fuzzer.cpp` — the harness used to trigger and reproduce the crash, calling `Program::load_method()` directly178 179## huntr Submission Note180 181Per the huntr MFV program's submission requirements, this PoC needs to be uploaded to a public HuggingFace repository before filing. `poc/poc_stride_int_overflow.pte` is ready for that upload.182