Overview

Truncation errors occur when a primitive is cast to a primitive of a smaller size and data is lost in the conversion.

Consequences

• Integrity: The true value of the data is lost and corrupted data is used.

Exposure period

• Implementation: Truncation errors almost exclusively occur at implementation time.

Platform

• Languages: C, C++, Assembly

• Operating platforms: All

Required resources

Any

Severity

Low

Likelihood of exploit

Low

Avoidance and mitigation

• Implementation: Ensure that no casts, implicit or explicit, take place that move from a larger size primitive or a smaller size primitive.

Discussion

When a primitive is cast to a smaller primitive, the high order bits of the large value are lost in the conversion, resulting in a non-sense value with no relation to the original value. This value may be required as an index into a buffer, a loop iterator, or simply necessary state data. In any case, the value cannot be trusted and the system will be in an undefined state.

While this method may be employed viably to isolate the low bits of a value, this usage is rare, and truncation usually implies that an implementation error has occurred.

Examples

This example, while not exploitable, shows the possible mangling of values associated with truncation errors:

#include <stdio.h>

int main() {   
  int     intPrimitive;    
  short   shortPrimitive;    

  intPrimitive = (int)(~((int)0) ^ (1 << (sizeof(int)*8-1)));    
  shortPrimitive = intPrimitive;    

  printf("Int MAXINT: %d\nShort MAXINT: %d\n", 
         intPrimitive, shortPrimitive);    
  return (0);
}

The above code, when compiled and run, returns the following output:

Int MAXINT: 2147483647
Short MAXINT: -1

A frequent paradigm for such a problem being exploitable is when the truncated value is used as an array index, which can happen implicitly when 64-bit values are used as indexes, as they are truncated to 32 bits.