5G flow-based datasets

This 5G dataset that can be used to advance research in attack detection in 5G service based architecture was generated using the open-source Free5GC testbed and UERANSIM, a UE/RAN simulator. The dataset include benign traffic featuring 5G procedures (e.g., registration, deregistration, PDU session establishment/release, uplink, downlink) along with 6 different HTTP/2 attack simulated between different 5G network functions. including:

• 2 variations of stream multiplexing attacks launched from SMF towards the AMF
• 1 rapid reset attack launched from PCF to UDR
• 3 variations of slow-rate attack launched from PCF to UDR

The dataset includes different CSV files (1CSV for benign data and 1 CSV for each attack simulation) depicting flow-based features extracted from the open source CICFLOWMETER.

We utilized CICFlowMeter to extract flow-based features from our 5GC dataset. CICFlowMeter, an open-source tool, generates bidirectional flows (Biflows) from PCAP files and extracts relevant features from these flows. This network traffic flow generator, creates bidirectional flows by determining the forward (source to destination) and backward (destination to source) directions based on the first packet observed.