CVE-2021-37639
published 2021-08-12CVE-2021-37639: TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can…
PriorityP336high7.8CVSS 3.1
AVLACLPRLUINSUCHIHAH
EPSS
0.17%
7.0th percentile
TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the list is empty this results in dereferencing a null pointer (undefined behavior). If, however, the list has some elements, if the restoration index is outside the bounds this results in heap OOB read. We have patched the issue in GitHub commit 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622. The fix will be included in TensorFlow 2.6.0. We will also cherrypick this commit on TensorFlow 2.5.1, TensorFlow 2.4.3, and TensorFlow 2.3.4, as these are also affected and still in supported range.
Affected
13 ranges
| Vendor | Product | Version range | Fixed in |
|---|---|---|---|
| debian | tensorflow | — | — |
| tensorflow | — | — | |
| tensorflow | — | — | |
| tensorflow | >= 2.3.0 < 2.3.4 | 2.3.4 | |
| tensorflow | >= 2.4.0 < 2.4.3 | 2.4.3 | |
| intel | optimization_for_tensorflow | >= 0 < 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622 | 9e82dce6e6bd1f36a57e08fa85af213e2b2f2622 |
| intel | optimization_for_tensorflow | >= 0 < 2.3.4 | 2.3.4 |
| intel | optimization_for_tensorflow | >= 2.3.0 < 2.3.4 | 2.3.4 |
| intel | optimization_for_tensorflow | >= 2.4.0 < 2.4.3 | 2.4.3 |
| intel | optimization_for_tensorflow | >= 2.5.0 < 2.5.1 | 2.5.1 |
| tensorflow | tensorflow | < 2.3.4 | 2.3.4 |
| tensorflow | tensorflow | — | — |
| tensorflow | tensorflow | — | — |
CVSS provenance
nvdv3.17.8HIGHCVSS:3.1/AV:L/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:H
nvdv2.04.6MEDIUMAV:L/AC:L/Au:N/C:P/I:P/A:P
vendor_debian8.4LOW
Stop checking back — get the weekly exploitation signal.
Every Monday: what got weaponized or added to CISA KEV in the last seven days — each CVE cross-linked to its PoC, Nuclei template, and detection rule. Free, one email a week, unsubscribe in one click.
Debian
CVE-2021-37639: tensorflow - TensorFlow is an end-to-end open source platform for machine learning. When rest...
vendor_debian·2021·CVSS 8.4
CVE-2021-37639 [HIGH] CVE-2021-37639: tensorflow - TensorFlow is an end-to-end open source platform for machine learning. When rest...
TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the
GHSA
Null pointer dereference and heap OOB read in operations restoring tensors
ghsa·2021-08-25
CVE-2021-37639 [HIGH] CWE-125 Null pointer dereference and heap OOB read in operations restoring tensors
Null pointer dereference and heap OOB read in operations restoring tensors
### Impact
When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer:
```python
import tensorflow as tf
tf.raw_ops.Restore(
file_pattern=['/tmp'],
tensor_name=[],
default_value=21,
dt=tf.int,
preferred_shard=1)
```
The same undefined behavior can be triggered by `tf.raw_ops.RestoreSlice`:
```python
import tensorflow as tf
tf.raw_ops.RestoreSlice(
file_pattern=['/tmp'],
tensor_name=[],
shape_and_slice='2',
dt=inp.array([tf.int]),
preferred_shard=1)
```
Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration:
```python
import tensorflow
OSV
Null pointer dereference and heap OOB read in operations restoring tensors
osv·2021-08-25
CVE-2021-37639 [HIGH] Null pointer dereference and heap OOB read in operations restoring tensors
Null pointer dereference and heap OOB read in operations restoring tensors
### Impact
When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer:
```python
import tensorflow as tf
tf.raw_ops.Restore(
file_pattern=['/tmp'],
tensor_name=[],
default_value=21,
dt=tf.int,
preferred_shard=1)
```
The same undefined behavior can be triggered by `tf.raw_ops.RestoreSlice`:
```python
import tensorflow as tf
tf.raw_ops.RestoreSlice(
file_pattern=['/tmp'],
tensor_name=[],
shape_and_slice='2',
dt=inp.array([tf.int]),
preferred_shard=1)
```
Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration:
```python
import tensorflow
OSV
CVE-2021-37639: TensorFlow is an end-to-end open source platform for machine learning
osv·2021-08-12
CVE-2021-37639 CVE-2021-37639: TensorFlow is an end-to-end open source platform for machine learning
TensorFlow is an end-to-end open source platform for machine learning. When restoring tensors via raw APIs, if the tensor name is not provided, TensorFlow can be tricked into dereferencing a null pointer. Alternatively, attackers can read memory outside the bounds of heap allocated data by providing some tensor names but not enough for a successful restoration. The [implementation](https://github.com/tensorflow/tensorflow/blob/47a06f40411a69c99f381495f490536972152ac0/tensorflow/core/kernels/save_restore_tensor.cc#L158-L159) retrieves the tensor list corresponding to the `tensor_name` user controlled input and immediately retrieves the tensor at the restoration index (controlled via `preferred_shard` argument). This occurs without validating that the provided list has enough values. If the
No detection rules found.
No public exploits indexed.
No writeups or analysis indexed.
https://github.com/tensorflow/tensorflow/commit/9e82dce6e6bd1f36a57e08fa85af213e2b2f2622https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gh6x-4whr-2qv4https://github.com/tensorflow/tensorflow/commit/9e82dce6e6bd1f36a57e08fa85af213e2b2f2622https://github.com/tensorflow/tensorflow/security/advisories/GHSA-gh6x-4whr-2qv4
2021-08-12
Published