CVE-2022-41887

TensorFlow crash: dimensions overflow in tf.keras.losses.poisson

• CVSS 7.5
• CWE-131 Incorrect Calculation of Buffer Size
• Numeric Errors
• Local

TensorFlow is an open source platform for machine learning. `tf.keras.losses.poisson` receives a `y_pred` and `y_true` that are passed through `functor::mul` in `BinaryOp`. If the resulting dimensions overflow an `int32`, TensorFlow will crash due to a size mismatch during broadcast assignment. We have patched the issue in GitHub commit c5b30379ba87cbe774b08ac50c1f6d36df4ebb7c. The fix will be included in TensorFlow 2.11. We will also cherrypick this commit on TensorFlow 2.10.1 and 2.9.3, as these are also affected and still in supported range. However, we will not cherrypick this commit into TensorFlow 2.8.x, as it depends on Eigen behavior that changed between 2.8 and 2.9.

CVE

CVE-2022-41887

CWE

CWE-131 Incorrect Calculation of Buffer Size

CVSS base score

7.5

Published

2022-11-18

OWASP

A06 Vulnerable and Outdated Components

Orthogonal defect classification

Function

Code defect classification

Incorrect Functionality

Category

Numeric Errors

Subcategory

Integer Overflows

Accessibility scope

Local

Impact

Denial of Service (DoS)

Affected component

TensorFlow

Fixed by upgrading

Yes

Solution

Update to TensorFlow version 2.11 or apply GitHub commit c5b30379ba87cbe774b08ac50c1f6d36df4ebb7c.

import tensorflow as tf

def poisson_loss(y_true, y_pred):
    return tf.keras.losses.poisson(y_true, y_pred)

y_true = tf.random.uniform((100000, 100000), minval=0, maxval=10)
y_pred = tf.random.uniform((100000, 100000), minval=0, maxval=10)

try:
    loss = poisson_loss(y_true, y_pred)
    print("Loss:", loss.numpy())
except Exception as e:
    print("Error:", e)

import tensorflow as tf

MAX_TENSOR_SIZE = 10000

def poisson_loss(y_true, y_pred):
    return tf.keras.losses.poisson(y_true, y_pred)

def validate_tensor_size(tensor):
    if tensor.shape[0] > MAX_TENSOR_SIZE or tensor.shape[1] > MAX_TENSOR_SIZE:
        raise ValueError(f"Tensor size too large: {tensor.shape}")
    
    return tensor

try:
    y_true = tf.random.uniform((100000, 100000), minval=0, maxval=10)
    y_pred = tf.random.uniform((100000, 100000), minval=0, maxval=10)

    y_true = validate_tensor_size(y_true)
    y_pred = validate_tensor_size(y_pred)

    loss = poisson_loss(y_true, y_pred)
    print("Loss:", loss.numpy())

except Exception as e:
    print("Error:", e)

@misc{vaitp:cve202241887,
  title        = {{TensorFlow crash: dimensions overflow in tf.keras.losses.poisson}},
  author       = {Bogaerts, Fr\'ed\'eric and Ivaki, Naghmeh and Fonseca, Jos\'e},
  year         = {2022},
  note         = {VAITP Python Vulnerability Dataset, entry CVE-2022-41887},
  howpublished = {\url{https://netpack.pt/vaitp/vulnerability/CVE-2022-41887/}}
}