I would tend to agree that sneaking past astral observers is quite possible, no, let me rephrase that, SHOULD BE quite possible.
Let's look at the modifiers:
Perception mods that don't get replaced by astral visibility:
Perceiver is distracted –2, for instance because the observer is travelling around at full astral walking speed, or because a small animal has its attention

Perceiver is actively looking/listening for it +3, possibly, but without an action, and not without having a hint that something might be there
Object/sound not in immediate vicinity –2, this counts on the astral as well
Object/sound far away –3, same
Object/sound stands out in some way +2, this one should apply for all auras of living things, and that's the full extent of the rules
Interfering sight/odor/sound –2, this one should be replaced by astral visibility
Perceiver has active enhancements +rating, not applicable
Perceiver using virtual reality –6, not applicable
Now astral visibility mods:
Sterile (clean room, hospital) +2, that should hardly apply in lots of places
Barren (city streets) +1, quite likely this one applies
Developed (suburban area, desert) 0, or this one
Cultivated (park, light forest) –1, in large compounds outside of cities, obviously
Teeming (jungle, forest) –2, more rare, but still...
Aura Noise
Devoid (no living traffic) +2, likely, but not definite
Sparse (scattered, occasional bystanders) +1, this might apply quite regularly
Moderate (frequent presence of living creatures) 0, I would say that's a rare facility to sneak into
Steady (regular movement of living creatures) –1, rare
Crowded (packed with living creatures) –2, again, in a jungle
Other Factors,
Shadow Clutter –1 to –4, this might apply quite often, for instance when crossing areas where stuff is loaded, stored, etc.
Background Count Inverse of Rating*, this one also: I would expect most industrial facilities to have at least a +1 BC.
Aerosol FAB cloud (p. 126) –2, well, if you put them there.
So, and finally, there are no listed mods for sneaking, so I would simply let an infiltrator roll his Infiltration, and apply the appropriate mods to perception.
A watcher with 2 regular assensing dice now gets, for an exemplary case, while sitting around only watching:
+0 because even while observing in detail, it doesn't know where to observe
- 2 if its watching a larger area
+2 because auras are obvious, (and I would apply another +2 for every spell or focus the infiltrator is sustaining)
+1 because it's probably a sparsely populated area, or else mundane stealth will get too hard
+1 because occasionally mundane security walks past
-1 to -2 because the infiltrator doesn't likely cross the empty causeway
-1 because there is a BC of +1. Oh, wait, now the watcher has force 0. Damn.
So let's take a Force 2 spirit, instead, and just use 2 dice as remaining from the force:
The resulting assensing dice are: 2 to 1. And Tadaa, the force 2 spirit is likely as not not going to beat the trained infiltrator at all, because he actually got 3-5 hits, statistically.
Now if I were lazy I would just forget all those mods, now that I've determined that for many normal cases in fact no modifiers apply to the assensing test. So... generally astral security will just roll their normal assensing dice, vs. the normal infiltration dice of the infilitrator. While this is still much worse than against mundane perception, it's also quite fair.