{"id":2816,"date":"2026-11-21T12:00:00","date_gmt":"2026-11-21T12:00:00","guid":{"rendered":"https:\/\/sites.wp.odu.edu\/locky\/?p=2816"},"modified":"2026-10-26T20:49:40","modified_gmt":"2026-10-26T20:49:40","slug":"neural-network-ids-vs-tree-ensembles-on-nsl-kdd","status":"publish","type":"post","link":"https:\/\/sites.wp.odu.edu\/locky\/2026\/11\/21\/neural-network-ids-vs-tree-ensembles-on-nsl-kdd\/","title":{"rendered":"Neural Network IDS vs. Tree Ensembles on NSL-KDD"},"content":{"rendered":"\n<p>I expected the neural network to win. It didn\u2019t.<\/p>\n\n\n\n<p>This assignment was a multi-layer perceptron (MLP) intrusion detector built in TensorFlow\/Keras on NSL-KDD network traffic. I used the same test set as my earlier Gradient Boosting assignment, so the comparison was direct.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Same detection, six times the alerts<\/h3>\n\n\n\n<p>The MLP missed 20 intrusions. Gradient Boosting missed 19, and Random Forest missed 18. On detection, that\u2019s a tie.<\/p>\n\n\n\n<p>False alarms were a different story: 27 for the MLP, against 5 and 4 for the trees. That\u2019s six times the alerts for the same detection rate. In a security operations center, it\u2019s the difference between a tool analysts trust and one they learn to ignore.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">More work for a worse result<\/h3>\n\n\n\n<p>The trees needed no feature scaling. The MLP needed log transforms, standardization, a validation split, early stopping, and a learning-rate search.<\/p>\n\n\n\n<p>That fits a broader pattern. Neural networks shine on data with structure, like images or sequences. Network connection records are a table of independent features, so there\u2019s no structure to exploit.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Where the errors lived<\/h3>\n\n\n\n<p>Over half the false alarms were ordinary ping traffic. In this dataset, ICMP echo also carries smurf attacks, and the features can\u2019t tell the two apart.<\/p>\n\n\n\n<p>Fifteen missed intrusions scored near zero. No threshold adjustment would recover them.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">An honest note on tuning<\/h3>\n\n\n\n<p>Tuning showed that learning rate mattered and batch size didn\u2019t. But the top four settings sat closer together than the noise from rerunning a single one. Tuning didn\u2019t produce a measurably better model.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">The takeaway<\/h3>\n\n\n\n<p>The newer model isn\u2019t automatically the better one. On tabular security data, the tree ensembles were simpler and more accurate.<\/p>\n\n\n\n<p>The full report is on my <a href=\"https:\/\/sites.wp.odu.edu\/locky\/coursework\/cybersecurity-applications\/cyse-420-applied-ml-in-cybersecurity\/\">CYSE 420 course page<\/a>.<\/p>\n\n\n\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\"><div class=\"wp-block-group__inner-container\">\n<div class=\"wp-block-columns is-layout-flex wp-container-core-columns-layout-1 wp-block-columns-is-layout-flex\">\n<div class=\"wp-block-column is-vertically-aligned-center is-layout-flow wp-block-column-is-layout-flow\" style=\"flex-basis:100%\">\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\"><div class=\"wp-block-group__inner-container\">\n<div class=\"wp-block-group is-layout-constrained wp-block-group-is-layout-constrained\"><div class=\"wp-block-group__inner-container\"><a href=\"https:\/\/sites.wp.odu.edu\/locky\/wp-content\/uploads\/sites\/38189\/2026\/10\/CYSE420_Assignment06_Report.pdf\" class=\"pdfemb-viewer\" style=\"\" data-width=\"max\" data-height=\"max\"  data-toolbar=\"both\" data-toolbar-fixed=\"on\">CYSE420_Assignment06_Report<br\/><\/a>\n<p class=\"wp-block-pdfemb-pdf-embedder-viewer\"><\/p>\n<\/div><\/div>\n<\/div><\/div>\n<\/div>\n<\/div>\n<\/div><\/div>\n\n\n\n<ul class=\"wp-block-social-links has-normal-icon-size has-visible-labels has-icon-color has-icon-background-color is-style-default is-horizontal is-content-justification-center is-layout-flex wp-container-core-social-links-layout-1 wp-block-social-links-is-layout-flex\"><li style=\"color: #ffffff; background-color: #cf2e2e; \" class=\"wp-social-link wp-social-link-chain has-white-color wp-block-social-link\"><a rel=\"https:\/\/www.credly.com\/users\/carl-lochstampfor-jr noopener nofollow\" target=\"_blank\" href=\"http:\/\/sites.wp.odu.edu\/locky\/wp-content\/uploads\/sites\/38189\/2026\/10\/CYSE420_Assignment06_Report.pdf\" class=\"wp-block-social-link-anchor\"><svg width=\"24\" height=\"24\" viewBox=\"0 0 24 24\" version=\"1.1\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" aria-hidden=\"true\" focusable=\"false\"><path d=\"M15.6,7.2H14v1.5h1.6c2,0,3.7,1.7,3.7,3.7s-1.7,3.7-3.7,3.7H14v1.5h1.6c2.8,0,5.2-2.3,5.2-5.2,0-2.9-2.3-5.2-5.2-5.2zM4.7,12.4c0-2,1.7-3.7,3.7-3.7H10V7.2H8.4c-2.9,0-5.2,2.3-5.2,5.2,0,2.9,2.3,5.2,5.2,5.2H10v-1.5H8.4c-2,0-3.7-1.7-3.7-3.7zm4.6.9h5.3v-1.5H9.3v1.5z\"><\/path><\/svg><span class=\"wp-block-social-link-label\">Full Screen or Download<\/span><\/a><\/li><\/ul>\n","protected":false},"excerpt":{"rendered":"<p>My Keras MLP matched Gradient Boosting and Random Forest on detection but raised six times the false alarms. What that says about tabular data.<\/p>\n","protected":false},"author":30379,"featured_media":2817,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"image","meta":{"footnotes":"","wds_primary_category":165},"categories":[165,60],"tags":[267,282,281,293,274,294,276,291,292],"_links":{"self":[{"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/posts\/2816"}],"collection":[{"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/users\/30379"}],"replies":[{"embeddable":true,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/comments?post=2816"}],"version-history":[{"count":1,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/posts\/2816\/revisions"}],"predecessor-version":[{"id":2818,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/posts\/2816\/revisions\/2818"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/media\/2817"}],"wp:attachment":[{"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/media?parent=2816"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/categories?post=2816"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/sites.wp.odu.edu\/locky\/wp-json\/wp\/v2\/tags?post=2816"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}