Folding Forum

 "In October 2017 we announced (https://developer.sonymobile.com/2017/1 ... en-source/) the open sourcing of the Folding@Home Android client developed by Sony.  We want to thank everyone who have contributed with CPU time. From the 16th of February 2018, it will no longer be possible to use the Folding@Home Android client from Sony." – Håkan Jonsson, Principal Engineer Sony Mobile Communications.
 
We want to thank Sony Mobile for help to take the Android client from an idea to reality.
We are now working on an overhaul of the backend before we relaunch the Android client. At this stage we do not have a date for relaunch.
But we will keep you posted.

I hope that this time, the Android client will be integrated to the usual infrastructure so that we can use it as a normal client with the same stats system.

Otherwise, I'll continue to ignore it as I used to do so ...

Yes would love to have proper integration with stats. I have years on the Android client that were not counted

I am in it for the science, so better utilization of CPUs, and perhaps, GPUs is my wish-list.

Better compatibility with different smart phones e.g. which have 4 fast CPU cores and 4 slow CPU cores.

If you develop some parts of the FAH app as open source on github then maybe volunteers could help with coding some features.

Kuno wrote:Yes would love to have proper integration with stats. I have years on the Android client that were not counted

Nice!! I've only been using it for a few months across a few android devices. It's so fun being part of a project that puts to use all of these latent tech resources. Can't wait for this new Android client to be released!!

My main concern / mind blocker is how an integration still could be open source without the risk of ... cheating ....let it be on the scientific code or just integration into stats. That might cause some addition measures in the backend. I might and hope to be wrong

Although admittedly I don't know the mechanism for the Android FAH client's security, the methods in many of the most important open source cryptocurrencies have often proven highly successful. Bitcoin, for example, is entirely open source; Ethereum, as well. When bugs are found in these open source initiatives, the community reacts. I believe that Pande Group would react to such findings in a manner to assure us that fairness continues to be of the highest priority. I believe that in their FAQs or in the Forum they have mentioned doing so before when having uncovered cheating (e.g. nullifying someone's score or banning a user). It would also drive innovation of new versions of the client toward greater security.

As currently part of the official stats-reported computer hardware client is closed source to reduce the risk of cheating, we are ultimately relying on the integrity of Pande Labs and their sponsor, Stanford University, as we participate in this project. I believe in their integrity.

It would also be possible to have the fah client open source but keep some core component closed source. This way FAH can hide some things they want to hide but still the other parts of the software can be improved by the public.

Hypocritus wrote:\I believe that in their FAQs or in the Forum they have mentioned doing so before when having uncovered cheating (e.g. nullifying someone's score or banning a user). It would also drive innovation of new versions of the client toward greater security.

The cheating scandals that I remember did not involve a compromise of the software security methology that's built-in to FAHClient and the servers. Punishment was exacted when results were returned due to some kind of virus-like code (obtained from websites that were unrelated to FAH's servers) which made unauthorized client installations without the knowledge or permission of the computer's owner.

Since AFAIK, the security methology has not been compromised -- and a genuine client was being run, the science was not compromised.

I would expect that the upcoming android client would use similar security methology, which probably includes portions of the client that may be ClosedSource.

SETI decided to combat fake results by requiring every WU to be run at least twice and compared, but that's an expensive waste of donated resources.

bruce wrote:
SETI decided to combat fake results by requiring every WU to be run at least twice and compared, but that's an expensive waste of donated resources.

Eh, that's not quite accurate.

BOINC (which SETI runs under) typically requires a minimum of 2 runs of the same WU. If they don't match, then another may be sent out. If they still don't, repeat to a preset max (think default is 5) before the server concludes something is wrong with the work.

The reasoning for that is because all kinds of things can go haywire on the user end. They could have bad hardware (overheating, bad RAM, etc.) to software issues (changes to the OS, other software esp. antiviruses), to people tinkering with the clients, but not necessarily maliciously. It's not always the user side though. Sometimes it's the client software or issues with the server.

If you think about how science works, you never just accept the first result... well, that's how it's SUPPOSE to work. You repeat the work to verify it works the same both times.

For cheating, it's really more about unauthorized use of computers. Years ago they discovered there was an individual who worked at a school district who had installed SETI on all the district computers and was appropriating systems and taking them home to run SETI. The community was highly suspicious of this individual. Eventually, that person was caught by the district, arrested, and prosecuted. His scores were wiped. There was some fingerpointing at Berkeley (SSL) who rightly pointed out that their license specifically pointed out it was for use only on systems the user had authorization to run them on. They had no control over where it was installed. Don't know if Folding ever experienced such a similar 'scandal.'

Quite some time ago FAH did experience one or more cases where an individual violated the EULA in a similar way. That did result in a zeroing of his account and a loss of his job. I don't think there has been a similar event recently.

As far as bad results are concerned (whether malicious aka "fake" or unintentional aka "corrupt") FAH is able to detect (almost?) all of them at upload time and they're rejected (zero points) and reassigned to someone else. Like BOINC, there's also a cut-off for the number of times a WU can be reassigned and when it's exceeded, that trajectory is suspended.

I have been assured that if a corrupt WU manages to get past those initial validation steps, it'll be obvious to the scientist and it won't be used for research. I'm not aware of that has actually happened, though.

My Android client keeps running and reporting statistics (I haven't actually confirmed CPU usage during nightly runs)
Does the announcement mean that even thought the client is on, it's merely waiting for a work unit assignment - or are there still mobile folding jobs in the pipeline? Thx.

wuffy68 wrote:My Android client keeps running and reporting statistics (I haven't actually confirmed CPU usage during nightly runs)
Does the announcement mean that even thought the client is on, it's merely waiting for a work unit assignment - or are there still mobile folding jobs in the pipeline? Thx.

Without having used the client, it sounds like the server may still be on, so technically yes the client is 'waiting' for work but since no work is being put into the server it may wait forever or until the server returns some error code or is unreachable.