Incorrect Behavior Order

Demonstrations

The following examples help to illustrate the nature of this weakness and describe methods or techniques which can be used to mitigate the risk.

Note that the examples here are by no means exhaustive and any given weakness may have many subtle varieties, each of which may require different detection methods or runtime controls.

Example One

The following code attempts to validate a given input path by checking it against an allowlist and then return the canonical path. In this specific case, the path is considered valid if it starts with the string "/safe_dir/".

String path = getInputPath();
if (path.startsWith("/safe_dir/"))
{
  File f = new File(path);
  return f.getCanonicalPath();
}

The problem with the above code is that the validation step occurs before canonicalization occurs. An attacker could provide an input path of "/safe_dir/../" that would pass the validation step. However, the canonicalization process sees the double dot as a traversal to the parent directory and hence when canonicized the path would become just "/".

To avoid this problem, validation should occur after canonicalization takes place. In this case canonicalization occurs during the initialization of the File object. The code below fixes the issue.

String path = getInputPath();
File f = new File(path);
if (f.getCanonicalPath().startsWith("/safe_dir/"))
{
  return f.getCanonicalPath();
}

Example Two

This function prints the contents of a specified file requested by a user.

function printFile($username,$filename){


  //read file into string
  $file = file_get_contents($filename);
  if ($file && isOwnerOf($username,$filename)){
    echo $file;
    return true;
  }
  else{
    echo 'You are not authorized to view this file';
  }
  return false;

}

This code first reads a specified file into memory, then prints the file if the user is authorized to see its contents. The read of the file into memory may be resource intensive and is unnecessary if the user is not allowed to see the file anyway.

Example Three

Assume that the module foo_bar implements a protected register. The register content is the asset. Only transactions made by user id (indicated by signal usr_id) 0x4 are allowed to modify the register contents. The signal grant_access is used to provide access.

module foo_bar(data_out, usr_id, data_in, clk, rst_n);
output reg [7:0] data_out;
input wire [2:0] usr_id;
input wire [7:0] data_in;
input wire clk, rst_n;
wire grant_access;
always @ (posedge clk or negedge rst_n)
begin

  if (!rst_n)

    data_out = 0;
  else
    data_out = (grant_access) ? data_in : data_out;
    assign grant_access = (usr_id == 3'h4) ? 1'b1 : 1'b0;


end
endmodule

This code uses Verilog blocking assignments for data_out and grant_access. Therefore, these assignments happen sequentially (i.e., data_out is updated to new value first, and grant_access is updated the next cycle) and not in parallel. Therefore, the asset data_out is allowed to be modified even before the access control check is complete and grant_access signal is set. Since grant_access does not have a reset value, it will be meta-stable and will randomly go to either 0 or 1.

Flipping the order of the assignment of data_out and grant_access should solve the problem. The correct snippet of code is shown below.

always @ (posedge clk or negedge rst_n)
begin

  if (!rst_n)

    data_out = 0;

  else

    assign grant_access = (usr_id == 3'h4) ? 1'b1 : 1'b0;
    data_out = (grant_access) ? data_in : data_out;


end
endmodule

See Also