Skip to content

An infinite loop in function cv::dnn::readNetFromTensorflow() #22709

@ConfZ

Description

@ConfZ

System Information

OpenCV version: 4.6.0
Operating System / Platform: Ubuntu 22.04
Compiler & compiler version: gcc-11.3.0

Detailed description

We find an infinite loop when calling the function "cv::dnn::readNetFromTensorflow(input_file)". When the function read our input_file, it hangs and cannot terminate. We found that it stuck in a loop (located in cv::dnn::dnn4_v20220524::RemoveIdentityOps line 833-840 ). The vulnerable code is listed as follows:

                while (true)
                {
                    IdentityOpsMap::iterator nextIt = identity_ops.find(it->second);
                    if (nextIt != identity_ops.end())
                        it = nextIt;
                    else
                        break;
                }

By our debug, the iterator "nextIt" displayed as "nextIt = {first = "", second = ""}" as the loop execution, namely "first = second" in the map iterator. Once the map identity_ops.find(it->second) executed, "nextIt" will cyclically assigned by "", by which the loop cannot terminate.

Steps to reproduce

To reproduce the bug, only need to call the function readNetFromTensorflow(repo) by our reproduce file . The reproduce file repro. We triggered the bug by the following simple code:

#include <opencv2/dnn/dnn.hpp>

using namespace cv;
using namespace dnn;

int main(int argc, char* argv[]){
  try {
    readNetFromTensorflow((const char*)argv[1]);
  } catch (std::exception &e) {}
return 1;
}

Issue submission checklist

  • I report the issue, it's not a question
  • I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
  • I updated to the latest OpenCV version and the issue is still there
  • There is reproducer code and related data files (videos, images, onnx, etc)

Metadata

Metadata

Assignees

Type

No type

Projects

No projects

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions