The　extensive　implementation　of　the　Internet　of　Things　(IoT)　is　vulnerable　to　security　and　privacy　factors　due　to　the　increased　use　of　Internet-enabled　devices.　The　data　without　interaction　between　Human-to-Human　or　Human-to-Computer　are　shared　by　the　IoT-enabled　devices　that　create　an　intelligent　system　of　systems.　The　data　capable　of　transforming　the　lives　of　humans,　businesses,　and　the　universe　are　extracted　by　the　IoT　systems.　But　in　a　highly　hostile　environment　such　as　the　Internet,　there　are　possibilities　of　an　end　number　of　cyber-attacks.　For　everyone,　including　consumers,　firms,　and　Government　Organizations,　the　primary　concern　is　to　protect　IoT.　However,　the　protection　of　the　systems　is　ineffective　due　to　the　complication　in　the　real-time　detection　of　the　attacks.　In　contrast,　the　complete　prevention　of　attacks　on　any　procedure　does　not　exist　forever.　The　research　on　competent　Deep　Learning　based　Intrusion　Detection　and　Preventions　Systems　(DL-IDPS)　conducive　to　IoT　environments　exists　in　limited　numbers.　IDPS　has　been　offered　numerous　DL-based　models　in　recent　years.　An　analogy　of　specific　deep　autoencoding　models　and　conventional　IDS　and　NIDS　datasets　has　been　proposed　in　this　research　paper.　Multi-Layer　Architecture-IoT　security,　the　Artificial　Neural　Networks　(ANNs)　are　deployed　for　this　IoT　research.　Each　layer　in　the　designed　architecture　has　been　assessed　with　ANNs　of　DL　on　the　KDD　Cup　’99　detection　of　intrusion　data　set.　The　present　technique’s　performance　results　over　the　KDD　Cup　’99　dataset　are　outperformed　by　this　innovative　research　with　97.77%　accuracy　and　0.71%　FAR.　©　2022,International　Journal　of　Network　Security.　All　Rights　Reserved.