Forum |  HardWare.fr | News | Articles | PC | Prix | S'identifier | S'inscrire | Aide | Shop Recherche
1964 connectés 

  FORUM HardWare.fr
  Seti et projets distribués

  Come back sur un article à propos de seti...

 



 Mot :   Pseudo :  
 
Bas de page
Auteur Sujet :

Come back sur un article à propos de seti...

n°491968
Seti@Donf
www.pronof1.fr
Posté le 06-09-2002 à 23:22:49  profilanswer
 

J'ai relu ceci :
http://www.wired.com/news/technolo [...] 38,00.html
 
et je trouve qu'on en est toujours au même point  :sarcastic:


---------------
Rejoignez la team HFR sur Pronof1.com (mot de passe : blabla) -> http://www.pronof1.com
mood
Publicité
Posté le 06-09-2002 à 23:22:49  profilanswer
 

n°491969
ienchangel
test
Posté le 06-09-2002 à 23:25:05  profilanswer
 

j'ai pas envie de tout lire, tu veux pas résumer stp  :D


---------------
Showire - Automatic Show Notifier
n°491970
Seti@Donf
www.pronof1.fr
Posté le 06-09-2002 à 23:30:18  profilanswer
 

ienchangel a écrit a écrit :

j'ai pas envie de tout lire, tu veux pas résumer stp  :D  




 
Faignasse  :na:


---------------
Rejoignez la team HFR sur Pronof1.com (mot de passe : blabla) -> http://www.pronof1.com
n°491971
ienchangel
test
Posté le 06-09-2002 à 23:37:03  profilanswer
 

Seti@Donf a écrit a écrit :

 
 
Faignasse  :na:  



t'as vu l'heure aussi?je me suis levé à 6 heures ce matin...je suis pas en étatt de lire tant que ça sur un écran d'ordinateur, parce que c'est pas confortable


---------------
Showire - Automatic Show Notifier
n°492110
ienchangel
test
Posté le 07-09-2002 à 11:59:41  profilanswer
 
n°492114
Raistlin M​ajere
i bouh at you !
Posté le 07-09-2002 à 12:04:15  profilanswer
 

SAN FRANCISCO -- Cheating to be top ET searcher on Seti@home may be more widespread than previously believed.  
 
The project's chief said that while distributed computing projects are a great way to accelerate data processing, peer-to-peer companies must be mindful of those who would compromise security and data integrity.
 
 
Seti@home, the biggest distributed computing project on the planet, is looking for signs of aliens in radio signals received from outer space.  
 
The project has recruited 2.7 million volunteers around the globe, who crunch the radio data on their home machines. To encourage participants, Seti@home ranks individuals according to the number of work units they've completed.  
 
Since its launch, the project has been plagued by a small number of people cheating to elevate their ranking in the leader boards.  
 
The Seti@home team has always dismissed cheating as a minor nuisance, but at this week's O'Reilly Peer-to-Peer conference, Seti@home's director, David Anderson, said security has been a major problem.  
 
"Fifty percent of the project's resources have been spent dealing with security problems," Anderson said during a talk entitled "Lessons from Seti@home."  
 
Anderson said some of the problems included dealing with a lack of a firewall at the University of Berkeley, where the project is based, but "the really hard part has to do with verifying computational results."  
 
Anderson said the Seti@home software had been hacked -? some were malicious, others not -? to make it run faster, to spoof positive results and to make it look more work had been performed to improve leader board rankings.  
 
He also said there were problems with over-clocked PCs, which tend to introduce errors into the calculations.  
 
Anderson told the audience that the hacking wasn't catastrophic because all the work units are sent out to two or more people, which acts as a checking mechanism, but he said it might be a more serious problem in other distributed computing tasks.  
 
As well as heading up Seti@home, Anderson is the CTO of United Devices, a startup planning to turn a profit from distributed computing.  
 
Anderson said that distributed computing volunteers have been most helpful to the Seti@home initiative. They have translated the website into 30 different languages, designed an attractive logo, and written software that farms out data to machines not directly connected to the Internet.  
 
"Users are eager to help," he said.  
 
The cheating was always dismissed as a nothing more than a nuisance, but one of the leaders of the biggest Seti teams suggested that cheating may be widespread.  
 
"It's (Seti@home's) biggest concern," said Ken "Caesar" Fisher of Ars Technica, an online technical magazine that has one of the biggest teams.  
 
Fisher said the latest version of the Seti@home software, Version 3.03, has been hacked to send back empty work units, elevating cheaters on the leader boards for units that haven't actually been analyzed.  
 
"It looks like they have a farm of Seti servers when all they have is an old 486," Fisher said.  
 
Fisher said the hacked software has been widely distributed through websites, chat sessions and e-mail.  
 
Fisher said the Ars Technica team recently discovered two, possibly three, cheaters in the ranks of its team, which has 3,800 members and is regularly among the top five most productive Seti@home squads.  
 
Fisher said the cheaters had set up a dozen separate accounts to cover their tracks.  
 
To make it look like the accounts were held by real people, the cheaters held conversations with themselves on the Ars' chat boards.  
 
"It's truly strange," Fisher said. "It's the behavior of someone who has serious social problems."  
 
The cheaters were discovered because the e-mail identities, which shared common IP addresses, suddenly rose to the top of the ranks.  
 
"These guys had tremendous scores," he said. "If someone is churning out 10 times as much as anyone else, there needs to be an explanation."  
 
Fisher said the cheating was unconscionable and he was angry that the Ars team had been exploited by the cheats.  
 
"It's hard to understand why, socially and psychologically, someone would do this," Fisher said. "I guess it's like being in high school and hanging out with the guy who can bench 350 pounds. It kinda makes you look cool."  
 
After his talk, Anderson downplayed the cheating.  
 
"It generates a lot of noise," he said. "But it's a really infrequent occurrence."  
 
Anderson said Seti@home had discovered and banned about 20 cheats. Falsified units are easy to detect, Anderson said, because they are empty. He estimated the percentage of falsified work units equaled less than one percent.  
 
However, Anderson said a security mechanism designed to detect cheating wasn't working. Whether that means cheating is more widespread, Anderson wasn't clear.  
 
"I don't think it's a widespread problem," he said.  
 
If people are spoofing Seti@home to enhance their self esteem, what are the chances they will cheat distributed computing projects that pay them for taking part?  
 
Nelson Minar, CTO and co-founder of Popular Power, another for-profit distributed computing company, said the same procedures for dealing with "Byzantine failures," or bugs, also work against cheats.  
 
"Protecting against bugs is the same as protecting against spoofs," he said. The methods range from running spot checks to sending out every calculation twice or more.  
 
Minar said its often useful to keep detailed stats. Anomalies like a sudden jump in the leader boards is a good indication of a problem.  
 
He said calculations sometimes include a checksum, or test calculation, that must be successfully tabulated.  
 
To prevent checksums being spoofed, he said experts are developing ways to take snapshots of the intermediary stages of a calculation.  
 
"It's a hard technique, but its hard to spoof" he said.  
 
Minar said the best protection comes from the architecture of a distributed computer itself: individuals can't choose what work it will do ?- they have to accept what they are given.  
 
He said encrypting the data sometimes helps, but it is not a fool-proof solution because the data is decrypted in memory as it s being worked on, making it vulnerable.

 
en gros : y a des tricheurs, c po bien :D


---------------
http://boinc.roadrunners.fr/  Feedback "l'habitat urbain satisfait toutes les femmes ; les sons que l'on débite avec la bouche"
n°492165
ienchangel
test
Posté le 07-09-2002 à 14:18:00  profilanswer
 

[citation]"It's truly strange," Fisher said. "It's the behavior of someone who has serious social problems."[/citation][:rotflmao]
 
de toute façon, tricher, au début, je trouve que c'est bien, qu'on est mieux classé, mais après je regrette, et je sais que je ne mérite pas cette place, et ça me ferait alors abandonner...
 
ps:j'ai jamais triché à seti, je sais pascomment fgaire, par contre, à dolphin, j'avais un peu cheater, mais après j'ai arrêté, pour les raisons citées ci-dessus


---------------
Showire - Automatic Show Notifier
n°492643
jaguargorg​one
I'm Jack's subliminal warning
Posté le 08-09-2002 à 13:25:08  profilanswer
 

y'avait un article dans New Scientist le mois dernier, qui expliquait que l'équation de Drake surrestimait beaucoup les probabilités de trouver du monde. Si je le retrouve je le scanne.


---------------
Un modérateur n'est pas là que pour emmerder le monde. Il a aussi un rôle important d'organisation de la section, un peu comme un CRS :o
n°492644
Benweb83
Profil: iwinter
Posté le 08-09-2002 à 13:26:03  profilanswer
 

Ch'ai pas si le Doc sera d'accord pour etre passé au scanner .....  :sarcastic:


Message édité par Benweb83 le 08-09-2002 à 13:26:10

---------------
HFRCoin: 7f1PV1XZc2W9848ykRFEwTuCZzchtTT83W -- Tu l'as trop écrasé, César, ce Port-Salut !

Aller à :
Ajouter une réponse
  FORUM HardWare.fr
  Seti et projets distribués

  Come back sur un article à propos de seti...

 

Sujets relatifs
[SETI] tps CPUSETI BIRTHDAY
Reinstallation de Setije viens de me mettre à SETI, et...
JOURNEE BLABLA SUR SETI.... Flood authorise jusqu'a 20hSETI, maths et shell
[SETI] Que pensez-vous de notre site ?Seti --> Expliquations SVP
[Classement SETI HFR] TOP 20[(Pas si) H.S. (que cela)] Victime de la Seti-dépendance ?
Plus de sujets relatifs à : Come back sur un article à propos de seti...


Copyright © 1997-2018 Hardware.fr SARL (Signaler un contenu illicite) / Groupe LDLC / Shop HFR