Abstract: Dengue is a mosquito-borne viral disease for which no specific antiviral drug is in routine clinical use, and the viral NS2B–NS3 serine protease, which processes the viral polyprotein, is an attractive target for direct-acting antivirals. Five curcumin-based compounds were evaluated with an integrated in silico workflow: curcumin (C1), desmethoxycurcumin (C2), bisdemethoxycurcumin (C3), a piperidone–sulfonyl curcumin analogue (C4) and a benzothiazole-linked curcumin hybrid (C5). Drug-likeness was assessed with Molinspiration and SwissADME, toxicity with ProTox-3.0, and the multi-target context of curcumin in dengue by network pharmacology. Binding to the NS2B–NS3 protease (PDB ID: 3L6P) was predicted by docking with PyRx/Auto Dock, and the C5 complex was examined in a 100 ns molecular dynamics simulation in GROMACS. Network analysis identified 37 targets shared between curcumin and dengue and ten host hub genes (IL1B, IFNG, CCL2, CXCL10, IL6, TLR4, IL10, IL17A, TNF and CSF2) enriched in cytokine and inflammatory signaling. Docking scores were −6.3, −7.2, −6.8, −9.2 and −9.8 kcal/mol for C1–C5, respectively. The two designed analogues scored best, and their predicted poses involved Asp75, Ser135 and His51 of the catalytic triad. In the 100 ns simulation the protease backbone remained stable (RMSD about 0.15–0.20 nm) and compact, but protein–ligand hydrogen bonds were transient. All results are computational predictions; C5 is regarded as a candidate for synthesis, enzymatic inhibition assays and further optimization rather than a confirmed inhibitor.