{"id":2162,"date":"2025-03-03T14:01:35","date_gmt":"2025-03-03T14:01:35","guid":{"rendered":"https:\/\/cybersecurityinfocus.com\/?p=2162"},"modified":"2025-03-03T14:01:35","modified_gmt":"2025-03-03T14:01:35","slug":"forscher-entdecken-llm-sicherheitsrisiko","status":"publish","type":"post","link":"https:\/\/cybersecurityinfocus.com\/?p=2162","title":{"rendered":"Forscher entdecken LLM-Sicherheitsrisiko"},"content":{"rendered":"
\n
\n
\n
\n
<\/div>\n
srcset=”https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?quality=50&strip=all 6000w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=300%2C168&quality=50&strip=all 300w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=768%2C432&quality=50&strip=all 768w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=1024%2C576&quality=50&strip=all 1024w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=1536%2C864&quality=50&strip=all 1536w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=2048%2C1152&quality=50&strip=all 2048w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=1240%2C697&quality=50&strip=all 1240w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=150%2C84&quality=50&strip=all 150w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=854%2C480&quality=50&strip=all 854w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=640%2C360&quality=50&strip=all 640w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=444%2C250&quality=50&strip=all 444w” width=”1024″ height=”576″ sizes=”(max-width: 1024px) 100vw, 1024px”>Forscher haben Anmeldeinformationen in den Trainingsdaten von Large Language Models entdeckt.\n

BOY ANTHONY \u2013 Shutterstock.com<\/p>\n<\/div>\n

Beliebte LLMs wie DeepSeek<\/a> werden mit <\/a>Common Crawl<\/a> trainiert, einem riesigen Datensatz mit Website-Informationen. Forscher von Truffle Security<\/a> haben k\u00fcrzlich einen Datensatz des Webarchives analysiert, der \u00fcber 250 Milliarden Seiten umfasst und Daten von 47,5 Millionen Hosts enth\u00e4lt. Dabei stellten sie fest, dass rund 12.000 hartcodierte Live-API-Schl\u00fcssel und Passw\u00f6rter dazu z\u00e4hlen.<\/p>\n

Der Analyse zufolge enth\u00e4lt das Datenpaket von Common Crawl insgesamt 219 verschiedene Geheimnistypen. Darunter Amazon Web Services (AWS) Root-Schl\u00fcssel, Slack-Webhooks und Mailchimp-API-Schl\u00fcssel. Da man sich mit diesen Anmeldedaten erfolgreich authentifizieren kann, stellen sie sowohl f\u00fcr Benutzer als auch f\u00fcr Organisationen ein erhebliches Sicherheitsrisiko dar.<\/p>\n

\u201eDies verdeutlicht ein wachsendes Problem: LLMs, die mit unsicherem Code trainiert wurden, k\u00f6nnen versehentlich unsichere Ausgaben generieren\u201c, warnen die Forscher. Demnach k\u00f6nnen LLMs w\u00e4hrend des Trainings nicht zwischen g\u00fcltigen und ung\u00fcltigen Geheimnissen unterscheiden. \u201eDas bedeutet, dass beide gleicherma\u00dfen zur Bereitstellung unsicherer Codebeispiele beitragen\u201c, hei\u00dft es im Forschungsbericht.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>","protected":false},"excerpt":{"rendered":"

srcset=”https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?quality=50&strip=all 6000w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=300%2C168&quality=50&strip=all 300w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=768%2C432&quality=50&strip=all 768w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=1024%2C576&quality=50&strip=all 1024w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=1536%2C864&quality=50&strip=all 1536w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=2048%2C1152&quality=50&strip=all 2048w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=1240%2C697&quality=50&strip=all 1240w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=150%2C84&quality=50&strip=all 150w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=854%2C480&quality=50&strip=all 854w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=640%2C360&quality=50&strip=all 640w, https:\/\/b2b-contenthub.com\/wp-content\/uploads\/2025\/03\/shutterstock_2450854311.jpg?resize=444%2C250&quality=50&strip=all 444w” width=”1024″ height=”576″ sizes=”(max-width: 1024px) 100vw, 1024px”>Forscher haben Anmeldeinformationen in den Trainingsdaten von Large Language Models entdeckt. BOY ANTHONY \u2013 Shutterstock.com Beliebte LLMs wie DeepSeek werden mit Common Crawl trainiert, einem riesigen Datensatz mit […]<\/p>\n","protected":false},"author":0,"featured_media":2163,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[],"class_list":["post-2162","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-education"],"_links":{"self":[{"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=\/wp\/v2\/posts\/2162"}],"collection":[{"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"replies":[{"embeddable":true,"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=2162"}],"version-history":[{"count":0,"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=\/wp\/v2\/posts\/2162\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=\/wp\/v2\/media\/2163"}],"wp:attachment":[{"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=2162"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=2162"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/cybersecurityinfocus.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=2162"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}