Agent Smith Problem, The problem and how to solve it Options

[Serbitar]

At the moment a hacker can have a large number of agents by just uploading infinite numbers to nodes and giving them orders.

Same goes for corporations, they could just have infinite numbers of agents.


Why I dont think it is a problem from the hacker side:

Stealth: At least for me, hacking is about stealth. An attack with infinite numbers of agents will instantly be noticed. And the hacking run is essentially over. Thats why I dont think the agent smith attack will be used, especially in very certain situation, here it might even make sense.

My problem is the corp version. At the moment I am handwaving it away by invoking "traffic" and "miss alarm" arguments.

In my own hacking rules I suggest dice modifiers if a persona/agent is subject to several interactions in the same phase. Be it perception or attacks, making large numbers of agents virtually useless.
I handwave this with the following arguments: Attacking and perceptions are active interactions with the persona to get data the persona doesnt really want to send (perception) or to force malevolent code into the persona. Both is easier recognized by the personas "health" and "checking" routines if large quantities of malevolent code is injected.

Please post your assessment of the whole Agent Smith problem. What is the problem for you, is it one, and how could one solve it.

[BishopMcQ]

As was discussed here, the Agent Smith problem has limitations.

Yes, it is possible to load Agents in several nodes, however:

1. Said nodes need to be within Signal range of where you intend to be hacking.
2. There is a limit on active users currently accessing a specific node, based on (System x 2)
3. The autonomy of Agents means that each must be individually called, taking actions.
4. If the Agents are not loaded directly onto the node, then they must hack their way through the firewall. (Note: Probing the target allows for a backdoor to be created which will be open for several hours or more at GM's discretion. This is not a problem unless the GM allows it.)
5. If you are loading Agents actively into a node, there is:
a) the cost in actions it takes you,
b) the limitation on the number of agents before response problems.
6. There are limitations on the teamwork tests even if the hacker has bypassed all previously stated hurdles.

[Serbitar]

[BishopMcQ]

1: If you can use the matrix as a relay, then please explain why two devices must be within the signal range of the weakest signal. One-way communication would be possible within the longer range, but for any kind of response or interactive nature, they would need to be within range of the shorter.

2: No you aren't killing the node, simply not allowing new people to access it until all users which were online at the moment of attack have been booted. Note, I don't agree with the rule, but rather acknowledging that it exists.

3: You can only order agents that are on your subscription list. If you remove an agent from your subscription list, it will continue on with its previous set instructions.

4: Again you run into the problem of subscription lists, teamwork rules, and signal.

[Serbitar]

1: exactly the same way how two cell phones cann talk to each other without being in their relative signal ranges.

2: Even better, you keep the corps IC out once youve flooded it with your own agents . . .

3-4: you didnt read my comment about mass ordering agents, did you? You can order (simple action) up to 12x2 agents per pass. Thats 12x3x2 agents per round, that should be enough.

[BishopMcQ]

"no you can order a pack of agents"

There wasn't a lot to read...where was your other comment?

[Serbitar]

That was the comment. You can order multiple agents at a time, if you give them the same order.

[FrankTrollman]

Remember that for this problem to be intractable it doesn't literally have to put up billions of agents (although in abstract there's no reason why it couldn't) - it merely has to put up so many Agents that differences in skill and awesomeness between diffeent Hackers fades into the statistical background noise.

12 Agents is enough and more than enough to make the problem plain. Being a Hacker so awesome that you roll six more dice is half as big a deal as having one more die per agent...

-Frank

[BishopMcQ]

I was simply reminding you limitation of the subscription list, which presuming you are connected to at least one node, means you are limited to 11 agents. Then you have to update your subscription list, at least a free per agent, possibly a simple to update the whole list.

The tactic is feasible, but the rate is slower.

[Serbitar]

You forget the agents you run on your comlink . . .

[djinni]

[Glayvin34]

The best Matrix security is definitely physical separation from the Matrix.

[Spike]

I would suggest this is obvious: If twelve agents are all on the same node, you now have twelve people using teh same hacked account, unless each agent hacked in seperately.

Twelve (thirteen) people all hacking in nearly simultaniously, but seperately, is NOISY. The IC WILL see it, and if the IC doesn't, the Sysop will certainly know what is going on. The more things you have sucking up processor speed the more likely it is for someone to see it and do something drastic. So they launch heavy duty IC... adn your horde kills it, or the system turns to goo because they sent in every available peice of IC... and you are out, because shutting down teh node until they can patch the exploit you used is infallible.

As for ordering 12 agents at teh same time to do somethign (trace, etc...), there would be a point of diminishing returns, as they all attempt the exact same actions in teh exact same sequence (being computer programs), I would, as a GM, probably make one roll for the entire goon squad of Smiths, representing their collective efforts.

At a minimum, however, I'd have analyse attempts rolled against each agent and hacker seperately, so each extra 'helper' you have would increase your chances of being caught.

One of these days I'll actually collate my argument against the credstick 'myth' so that I can stop just handwaving it away as silly.

[ShadowDragon8685]

A Goon Squad of Agent Smiths would be an effective way to deal with enemy hackers/Big Bad IC, though.

You know, like the Burly Brawl, but the other guy dosen't get to be Neo. :)

[Thain]

The perfect solution for the Agent Smith Problem is for the gamemaster to exercisesome common sense, and say "No." if itdoes indeed become a problem.

[Glayvin34]

And Serbitar mentioned that it's also exclusively for cybercombat, after the jig is up.
You can also use your Agent Smiths to use Medic on your icon (since you've got Wired Reflexes 3 active and you're fighting in AR, right? right? haha :S ). And you can have your Agents ready loaded on your credstick ready to jump out and stomp a pickpocket or whoever you tell them to, or they tell each other. Sure, that requires an action on the Agent's part, but daisy-chain them and you've got 300 Spartans.

[Garrowolf]

well this was why I was trying to point out in the other thread that the system can perceive you and doesn't need the agents in the first place. The system will use IC when someone has touched a sensitive area but not to deal with normal hacking attempts.

One thing I was thinking about was what if you treated agents as processes on your commlink or system. Basically get rid of their ability to copy themselves onto another system. they don't need to move to datasearch. They can browse just like you can. It would work for most functions but it would get rid of the part that is a problem - hacking.

Make them just an interface tool and the whole issue goes away.

[FrankTrollman]

[ludomastro]

My solution for the Agent Smith problem:

[Thanee]

Allow only one copy of each Agent to work at a time... at least at the same node. ;)

Bye
Thanee

[mfb]

[Synner]

[mfb]

yeah, but you can bypass the subscription limit by simply subscribing commlinks to your commlink. the subscribed commlinks have agents subscribed to them (or more commlinks subscribed to more agents, or however deep you want to go). i believe that, using the rules for holding your action, if you pass an order to your army of network-subscribed agents, they will all execute it that same turn. even if not, that just means that it takes time for your army to mobilize. lots of time--maybe as much as an entire minute.

[Trigger]

I have find it extremely hilarious that this thread is running at the same time as the one 'How to Speed up Hacking', with many of the same people posting in each :grinbig:

[Synner]

Subscribing a commlink to your commlink does not automatically equate to subscribing software running on that commlink to yours. ie. you can't run Analyze stored on a subscribed commlink as if you had it uploaded, why should you be able to command an agent? You are assuming you can route batch commands to the agents somehow, but if that were the intention you could subscribe agents as groups on your commlink in the first place.