The following answer is not "rigourous", but it may gives a simple explanation.
Suppose you have a quantum fluctuation just near the horizon, but outside.
This quantum fluctuation create 2 particles, one with a negative energy -E, the other with a positive energy +E.
If the 2 particles stay outside the black hole, they have to anihilate themselves in a time time $t \le \frac{\hbar}{E}$
Now, one of the 2 particles may fall inside the black hole, and there are 2 possibilities ; the escaping particle may have a positive or a negative energy.
The key point is that there is an asymmetry between these 2 cases.
For a particle to be real, its energy has to be positive, but relatively to the time coordinate. With an evolution variable $\tau$, this can be write $\frac{dt}{d\tau}>0$
When the horizon is being crossed (by the infalling particle), we may consider, that there is a change in the nature of the time and radial space coordinate. The time-like coordinate becomes a space-like coordinate, and the radial space coordinate becomes a time-like coordinate.
More precisely, if, outside the black hole, the coordinate are(in units $c=1$) : $z=r+it$, then the "coordinates" inside the black hole are $z \rightarrow z'\sim -iz$
So, $$z'=r'+it'\sim-i(r+it)=(t-ir)$$
So, $t'\sim-r$ and $x'\sim r$
For an escaping particle, the energy must be positive relatively to $t$, so $ E=\frac{dt}{d\tau} >0$, but for the infalling particle the "energy" must be positive relatively to $t'$, that is = $\frac{dt'}{d\tau}>0$, which is "equivalent" to $-\frac{dr}{d\tau}>0$.
But the last expression means only that the particle is an infalling particle, which was our hypothesis. We could say also, for the infalling particle, that the "outside" energy $-E$ becomes a "inside" momentum.