Another proposal is just autoban or issue warning based on accuracy? We have data from the pro players and we know what the best of the best can do. Anyone who have better aim than the pros should get a few warnings then autoban.
This would limit the cheats "max skill" on the cost of the max being set in stone, also if poorly implemented it would lead to false positives which is a problem with all your solutions especially since there is no method to prove your innocence.
The games know where everyone is and what they are looking at. So if another player is snapping to targets repeatedly through walls that they clearly wouldnt know the position of autoban. Its one thing if they look toward an area they hear footsets or gun fire, but its another thing when they are having a hard time shooting another player because they keep snapping to another target behind an enemy they are engaging.
Also possible to implement to some extent and in your example it is clear cut, however same as previously this would only shrink obvious cheats. Since how do you determine if they heard the footsteps or not?
Why is dealing with cheating software so difficult? why isnt there a team automatically downloading the software from cheating websites and just reverse engineering it? The software sells, and eventually people get banned, and just create a new account immediately. Eventually there is a fix for that software and the cheaters just update to get around it and it starts all over again. Why isnt there someone just spending 20 bucks immediately the date the update is released and creating a fix?
You do not have full control over the system which runs the game and the input, and there is a large variety of actions/variables which is hard to distinguish legitimate and non legitimate. And In an arms race the cost is to much to get to the fine grain required unless of course you want to play something with very limited set of variables.
Of course you could increase the servers control of the hardware aka stadia(which would increase the difficult of cheating since you no longer can get the clients data to parse from and now would need to parse from the image/sound itself, but this would only slow it down until the individual get the processing power/code to do so) But that would to give even more control to corporations which as you can see is not used in your best interest.
Thoughts? Other than "git gud, noob" tolling?
Learn to code?
My suggestion is a return to the old days where the servers could be in the control of the community rather than the corporations due to the fact they could then implement the granularity of control and checks they want without us requiring to give up our own control of hardware. So you could have the casual servers which has lenient cheat control/moderation and the more esport/hardcore which would perhaps mostly be run in lan matches. (of course this is a pipe dream since corporation will never give up their control due to microtransactions).
Not sure you can in entirety dismiss the query to a yes or no in regards to if there are methods to distinguish a human input from a computer input, since that is obviously possible, the doubt comes into how large margin of error you allow.
But how would you go ahead and create this algorithm which computes input and then classified them as valid or non valid,First you would need to gather a large database of input in the format of whatever the game uses, This input would then need to be classified into series of inputs(the data need to be over time due to a classification on each individual input would not catch anything except for input that is out of bounds and give a lot of false positives) (This step would make it probably more difficult to use data from other games due to even though fps are similar to each other there are sequences which are fair in others (e.g aimbot exist in overwatch as part of the soldiers gameplay)) thus the data itself need to be prepared for each game. You then need to also gather data for a subset of different hacks and how they affect players input and generate some data from this, then the algo needs to be trained on this data ( my knowledge is not up to date atm on how to build the algo itself with the latest techniques) and it's result tested until you at least get 90% accuracy, once this is done you then need to implement the system to run the trained algo and here comes a new problem, you do not want this algorithm to affect performance of the game (we are already screwed with communications over the internet degrading performance) so this would undoubtedly stop it from running at the same time as the match is going (might work if there are low amount of player input to consider to have it done in realtime e.g 5 vs 5 or 3 vs 3) which means you need to save all the match data and then run it after the match is over even then this solution will still give false positive especially at the start of its deployment when you have minimum of data thus getting to my point of having a process of appealing a ban (Most of these security software in gaming works by having their process hidden in how they function in attempt to stop people from learning how to bypass them too quickly which would limit appeal since it would be the corporation itself judging if they are in right or wrong and simple cases such as your brother would be fine (unless they would find that it is more profit to ban people and make them buy the game again and again) but the hard to define would most likely be defined in favor of the corporation, which I would be fine with if there were a option to run my own server of the game thus the ban would only limit my ability to play on the corporation server but not the game itself)
The above example would cost quite a bit to implement and since it would most likely only be good for one game and perhaps even only one version of the game and if the game has a low entry cost such as apex then it would only train cheat maker in making a better software thus we get into the idea of an arms race, which I think current corporation do not find to be profitable at the moment.
The current software for driving a car down the road is limited to some extent when it comes to handle things in realtime and with other human drivers and pedestrians thus can mostly be classified as assisted driving rather than self driving in that it need a human operator on standby to handle more difficult environments.
Now to my very limited understanding of current anti-cheat software (due to their own secrecy) to some extent they use machine learning and algorithms but they mainly uses them in order to try and determine if the extra process you are running is a legit process or not similar to how anti-virus software try to match signatures of possible viruses running on a computer the problem being is you can quite quickly build a virus with some intended effect and then check with all the anti-virus vendor if they catch it and if they do try to obscure the process so it no longer matches their signatures thus the defender always have to react to the attacker meaning that they will most likely be one step behind.
So once more the optimal solution is to minimize your liberty even more and give the corporation full control since it is the most cost efficient method for them in order to remove all cheating, while my solution of having more decentralized will only limit the extent of cheating by making people naturally group together with non cheaters over time but not remove it in its entirety.