In most systems and standard libraries, malloc is optimized by allocating a larger chunk of memory than required (up to 64K on some systems) from the OS, and then internally disbursing it as per requirement from this pool. The same applied to the deallocation (free) as well, in that, the free'd memory isn't freed, but put back in the memory pool, in case another request comes in, in which case the pool is reused.
Thus, the call to free hasn't actually freed the memory from the process's address space and is thus accessible even after free.
My understanding as to why this is done is that system calls are expensive, so the initial call to malloc will make a system call for enough memory that future requests for malloc do not immediately trigger another system call for more memory.
As for further reading, please take a look at this page on Wikipedia, How do malloc() and free() work? and How is malloc() implemented internally?