La Double Inconstance (2030–31) — Episode 14
27: Engagement: the Audit-Core
The engineers who designed the module were occupationally paranoid, as all weapon systems designers have to be. Few assumptions survive engagement with the enemy: caution and redundancy are their watchwords.
The unchallenged leaders in paranoia are the security engineers. Their trade assumes that hostile entities are actively trying to break our systems. For such people the worst case is their starting assumption.
The audit-core of the module was sealed inside a tamper-proof package encapsulated within thermal armour. It was well-connected to the ship’s systems while its code was read-only. The system was data-filled and configured on installation back on Earth and could only be changed through a major refit. In all this it resembled an aircraft ‘black box’ flight recorder.
Its responsibility was to verify the module’s overall integrity. In case of serious doubt it would either withdraw the ship from danger if it could - or blow it up if it had to. It had a bias towards crew survival: but that was not its highest priority.
And now it was suspicious.
Unlike its biological analogues, the audit-core was not a simple, primitive, situation-action reflex system. Too much hung on its actions. Instead it had one of the most powerful classification and interpretation systems on the ship.
Right now it was worrying about the destruction of the mother-ship.
It seemed clear that the mother-ship had in fact been destroyed. Multiple sensors - radar, infrared, optical and heartbeat - told a consistent story. But the destruction event itself was puzzling.
The mother-ship’s own radar-view, transmitted in real-time, showed a compelling narrative of incoming kill-missiles. Yet the signal processing unit on the module itself, which had received this downlink traffic, had logged anomalous packet activity during the relevant sequence. It was not obvious - but it was there if you looked for it. This was consistent with a false signal injection from somewhere on the module itself.
Now the audit-core re-examined the explosion itself through a spectral analyzer, running a DC transform and comparing it with the spectrum of a reference self-destruct sequence. The results showed unusual high-frequency artefacts. Another peculiarity was that there was little evidence of a jet of plasma emerging opposite the presumed direction of impact, something conservation of momentum demanded. If anything, the explosion-pattern looked almost cartoonishly symmetric.
Its suspicions increased.
The audit-core was running a database of competing hypotheses using Bayesian updating. The chances that the destruction of the mother-ship had been misrepresented presently stood at 32%.
The probability that this might be due to a malicious signal from the module itself, injecting a false self-destruct command, was currently languishing at only 6% (other hypotheses were accident, systems failure and sabotage-malware in the mother-ship itself).
It was at this point that the module’s five nuclear missiles, each now dispersed thousands of kilometres from their launch site, blew up.
The audit-core, running on the fastest platform money could buy, was thinking one hundred thousand times more quickly than a human. It observed how - with glacial slowness - the module came around to combat alert status again. Meanwhile it reviewed the narrative the module’s sensors were providing - each nuclear missile hit by a high-velocity kinetic weapon at insane velocities. But after review, each piece of impact-imagery seemed to the audit-core as suspect as the last.
It ran a sub-thread to check the data from the first reconnaissance operation, the one where three drones had been eliminated, expecting to find evidence of fabrication there too. Meanwhile its high-level decision routines reviewed the evidence.
Hypothesis 1: the Adversary had hit the drones, the mother-ship and now the loiter-missiles. It was evidently clearing ambient space of mission assets. Logically, the module would be next. Indeed, impact vehicles were probably already in flight. Action: report back and then retreat as fast as possible.
Hypothesis 2: there had been no physical attacks at all. All the losses had been achieved through sabotage. This might be due to the Adversary; or to some other entity hiding behind the concept of Adversary. In any event, systems were corrupted and compromised from top to bottom. The module could no longer be trusted in any regard. The best option was: report back then self-destruct.
There were other possible hypotheses but these were the leading candidates, the ones with the highest probabilities. Neither could be rejected out of hand. The conservative option was to provisionally adopt hypothesis one: try to outrun any imminent attack. After all, option 2 could be fully actioned at any time.
But the mother-ship attackers had demonstrated a terminal acceleration of 35 g.
That was unfortunate.
For security reasons, the audit-core had not been provided with the disposition of any other mission forces. In case of escape-override from Martian orbit it had been provided with the coordinates of a destination some 1.2 million kilometres back from the planet. It had no need or ability to speculate but it would be logical to assume some asset had been positioned there to recover whatever was left.
At the sustained 40 gee which the module’s engines could pull, this would take under an hour.
28: ‘Secure for ultra-high acceleration’
Three milliseconds had passed since the module detected the first nuclear missile destruction event. The audit-core determined that the situation was now a Mission Emergency. It therefore initiated a command-override interrupt and assumed control, displacing the ship’s AI and the crew from the control loop.
It sent the crew and the models a terse command: secure for ultra-high acceleration and set an acceleration timer. Eight seconds. Destruction by the Adversary and self-destruction were much the same. The only real point of an escape-and-evasion option was to retrieve the crew plus the module itself for forensic analysis. Eight seconds seemed a reasonable compromise.
The audit-core began to assemble a summary of its findings together with a core-dump of the AI, preparing to transmit via a secure link of its own.
It did not have the capability to wonder whether the putative receiving system might also be corrupted.
Roll back time a few seconds, to when the audit-core had just begun its investigation, just started to indulge its concerns over data and traffic anomalies.
Tania is working out her twinges in the gym. She hates the hours in the combat-couch, wrapped in warm gel, unable to move and smothered by equipment. She wants to be free, to use her limbs, exhaust herself and do something. She is strapped to the treadmill, keeping up a fast pace, counting the kilometres.
André and Flaminia are engaged in an altogether different form of recreation.
They’re talking: inconsequential memories and reminiscences. André feels there are many personal facts and feelings he wants to share. You can tell if you really relate to someone, he thinks, if sharing feels natural, not just the dry recital of biographical data. She listens amiably, asks questions, is interested and engaged.
And then the lights begin to strobe and a deep, almost slowed-down bass voice fills the room: “S_e_c_u_r_e _ f_o_r_ …”
Flaminia reaches out and thrusts him towards the combat couch. The recoil sends her into a wall from where she bounces towards the gym - where Tania has gotten tangled in her restraining straps.
The countdown is at four seconds when Tania, thrown by Flaminia with astonishing strength into the office, grabs a bar on the wall and spins towards her couch. For Flaminia there is no more time: she rebounds towards the nearest cabinet where the autostrap system waits to bind her tightly as she enters module-virtuality.
With two seconds to go André is stowed in his combat couch, lid down and all support systems in place. Tania might have made it if the module hadn’t been pointing in the wrong direction. The initial acceleration is modest - just over one gravity - but the ship is turning at the same time and the shifting acceleration makes her miss her couch. Tania hits it awkwardly, stumbling and falling to her knees. She gets her feet to the floor and tries to push herself up. At two gee this is difficult and at three impossible. Tania must guess the outcome, but professional to the last, she lets herself crumple to lie as flat as she can on the carpet... as the weight of an elephant collapses her lungs.
The gee-forces keep climbing. Humans can survive hundreds of gee for tiny instants if properly protected; people can function at twenty gee for ten seconds or so (no-one volunteers twice for this) - while fighter pilots endure ten gees sustained as part of their job.
Forty gee is not survivable on an unprotected surface. Tania’s internal organs spread and rupture; her skull cracks under the weight of her brain which puddles under the enormous pressure. This is what kills her - before the drawn-out agonies of asphyxiation.
André is in a combat couch designed for any number of truly bad outcomes. Enormous counter-pressures are applied to his body (the couch is a remote descendant of the famous ‘g-suits’). Intense magnetic fields fight the pressures on his brain. He is minimally-conscious on this wild ride, oxygen and circulation provided only by machines.
He survives the first moments of forty gee: just.
As the module ascends on its blinding fusion flare, the audit-core watches through the module’s untrusted eyes, waiting for credible evidence of high-gee pursuers.
The full story text can be found in my SF novel: here:
- "Donatien's Children" (2022) — as a PDF, and
- "Donatien's Children" (2022) — on Amazon for easier reading.
No comments:
Post a Comment
Comments are moderated. Keep it polite and no gratuitous links to your business website - we're not a billboard here.