{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/aloha-auxiliary-loss-optimization-for","title":"ALOHA: Auxiliary Loss Optimization for Hypothesis Augmentation","arxiv_id":"1903.05700","date":"2019-03-13","proceeding":null,"authors":["Ethan M. Rudd","Felipe N. Ducau","Cody Wild","Konstantin Berlin","Richard Harang"],"abstract":"Malware detection is a popular application of Machine Learning for\nInformation Security (ML-Sec), in which an ML classifier is trained to predict\nwhether a given file is malware or benignware. Parameters of this classifier\nare typically optimized such that outputs from the model over a set of input\nsamples most closely match the samples' true malicious/benign (1/0) target\nlabels. However, there are often a number of other sources of contextual\nmetadata for each malware sample, beyond an aggregate malicious/benign label,\nincluding multiple labeling sources and malware type information (e.g.,\nransomware, trojan, etc.), which we can feed to the classifier as auxiliary\nprediction targets. In this work, we fit deep neural networks to multiple\nadditional targets derived from metadata in a threat intelligence feed for\nPortable Executable (PE) malware and benignware, including a multi-source\nmalicious/benign loss, a count loss on multi-source detections, and a semantic\nmalware attribute tag loss. We find that incorporating multiple auxiliary loss\nterms yields a marked improvement in performance on the main detection task. We\nalso demonstrate that these gains likely stem from a more informed neural\nnetwork representation and are not due to a regularization artifact of\nmulti-target learning. Our auxiliary loss architecture yields a significant\nreduction in detection error rate (false negatives) of 42.6% at a false\npositive rate (FPR) of $10^{-3}$ when compared to a similar model with only one\ntarget, and a decrease of 53.8% at $10^{-5}$ FPR.","url_abs":"http://arxiv.org/abs/1903.05700v1","url_pdf":"http://arxiv.org/pdf/1903.05700v1.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"aloha-auxiliary-loss-optimization-for","repo_url":"https://github.com/cmikke97/Automatic-Malware-Signature-Generation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"attribute","task_name":"Attribute"},{"task_slug":"malware-detection","task_name":"Malware Detection"},{"task_slug":"tag","task_name":"TAG"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}