Open Access
Table 1
Generic comparison of prior systems with proposed system.
| Feature | What prior systems does | Proposed system |
|---|---|---|
| Packet Queuing Strategy | Queuing is mostly rule-based or static; suspicious traffic is handled after table-miss events or via fixed thresholds; the queue itself is not steered by a learned model. | A controller resident queuing design that separates normal and suspicious flows before classifying, cut down unnecessary packet drop and there is a time limited blacklist that automatically unblocks. |
| Any Feature Selection Method | Feature sets are either fixed (hand-engineered or filter-only) or not adjusted for the specific classifier | Wrapper Based forward feature selection process for choosing compact features that makes the controller more accurate and less time consuming for performance. |
| SDN Integration | Detection often runs off the controller (as a separate module or service) and is loosely connected to mitigation, and this raises delays and controlling impact. | Ryu Controller: packet filtration to queuing mechanism and then classification from alert to block/unblock, which will simplify all the operations. |
| System Architecture | Detection and other parts are divided among components (like neighbor-switch detours, caches, and proxies) and may need changes to the device or protocol. | Unified pipeline with low overhead on SDN; has a reconnect policy so that benign hosts can start up again when things go back to normal. A cohesive pipeline cuts down on false positives and control plane load while making things better for real users. |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.