Isistimu yokuhlunga ogaxekile ye-Rspamd 3.9 iyatholakala

I-Rspamd 3.9, isistimu yokuhlunga ogaxekile, isikhishiwe. Ihlinzeka ngamathuluzi okuhlola imilayezo ngokusekelwe kumibandela ehlukahlukene, okuhlanganisa imithetho, izindlela zezibalo, nohlu oluvinjelwe, olusetshenziselwa ukunquma isisindo sokugcina somlayezo esisetshenziselwa ukunquma ukuthi kuvinjwe yini. I-Rspamd isekela cishe zonke izici ezisetshenziswe ku-SpamAssassin futhi inezici eziningi eziyivumela ukuthi ihlunge i-imeyili ngokwesilinganiso izikhathi ezingu-10 ngokushesha kune-SpamAssassin, futhi inikeze ikhwalithi yokuhlunga engcono. Ikhodi yesistimu ibhalwe ngo-C futhi isatshalaliswa ngaphansi kwelayisensi ye-Apache 2.0.

I-Rspamd yakhiwe kusetshenziswa ukwakheka okuqhutshwa yimicimbi futhi yenzelwe izinhlelo ezithwala umthwalo omkhulu, icubungula amakhulu emiyalezo ngomzuzwana. Imithetho yokuthola ogaxekile iguquguquka kakhulu futhi, ngendlela yayo elula, ingaqukatha izinkulumo ezijwayelekile, kanti ezimweni eziyinkimbinkimbi kakhulu, ingabhalwa nge-Lua. Ukusebenza kungandiswa futhi izinhlobo ezintsha zokuhlola zingangezwa ngamamojula angabhalwa nge-C kanye ne-Lua. Isibonelo, amamojula ayatholakala ukuze kuqinisekiswe umthumeli kusetshenziswa i-SPF, ukuqinisekiswa, nezinye izindlela. isizinda Ukuqinisekiswa komthumeli nge-DKIM, okudala imibuzo ohlwini lwe-DNSBL. Kunikezwa isikhombimsebenzisi sewebhu sokuphatha ukuze kube lula ukusetha, ukudalwa kwemithetho, kanye nokulandelela izibalo.

Enguqulweni entsha:

• Izilungiselelo ze-Bayesian classifier zithuthukisiwe. Usayizi wewindi ozenzakalelayo uncishisiwe usuka kumagama angu-5 ukuya kwangu-2, uthuthukisa ukusebenza nokunciphisa ukusetshenziswa kwesikhala sesitoreji ngesici esingu-4 ngaphandle kokuphazamisa ukusebenza kokuhlelwa kogaxekile. Insiza ye-"rspamadm classifier_test" iyatholakala ukuze kuhlolwe isihlukanisi ngezilungiselelo ezihlukile.
• Imojula ye-GPT yengeziwe, kusetshenziswa i-OpenAI GPT API yokuhlukanisa umbhalo ngokubuza amamodeli olimi amakhulu njenge-GPT-3.5 Turbo ne-GPT-4o. Ukunemba kokuhlelwa kogaxekile kwemojula entsha kungaphansi kwalokho kwesigaba sase-Bayesia, kodwa inzuzo yakhona ukuthi ayidingi ukuqeqeshwa kwangaphambili futhi ingacabangela umongo wemilayezo, kuyilapho isihlungi se-Bayesia sidinga ukuqeqeshwa kwenjini yekhwalithi ephezulu nokunokulinganisela ukuze kuphumelele. Ngokungeziwe ekukhombeni ngokuqondile ugaxekile emilayezweni, imojula ye-GPT ingasetshenziswa ukuqeqesha isigaba se-Bayesia.
• Ikhono lokusebenzisa amamojula aziwa_abathumeli nezimpendulo ndawonye liye lasetshenziswa ukuze kumake abathumeli abaqinisekisiwe, kusetshenziswa iqiniso lokuthi izimpendulo bezithunyelwe kubo ngaphambili njengenkomba.
• По умолчанию отключено динамическое изменение ограничений интенсивности отправки сообщений (dynamic ratelimit), связанных с одним отправителем, получателем или Ikheli le-IP.

Source: opennet.ru