Packet Log Analytics
Introduction
Packet logs provide a wire-level view of network activity. Each packet captures source/destination addresses, ports, protocol, timing, sizes and flags—enough detail to reconstruct conversations, baseline normal traffic, and catch abnormalities in real time. Conatix uses deep-learning models to learn those baselines and surface anomalies such as scans, floods, unusual protocol mixes, and lateral movement—without relying on brittle rule sets.
What We Analyze
- Packet volumes and burst patterns across time windows.
- Protocol distribution shifts (e.g., UDP spikes vs. TCP data/connect events).
- TCP flag irregularities (SYN, ACK, RST, FIN) that indicate scans or resets.
- 5-tuple flow behaviors (source/destination IP/port + protocol) and deviations from baselines.
- Latency/throughput outliers, fragmentation, and retransmission indicators.
Common Terms (Quick Reference)
- Packet: The smallest unit of network transmission; analyzed to detect fine-grained anomalies.
- Flow (5-tuple): Grouping of packets by src IP, dst IP, src port, dst port, protocol; core to behavior tracking.
- Protocol: Communication rules (e.g., TCP, UDP, ICMP). Different protocols imply different risks and performance traits.
- TCP Flags: Bits that describe connection state (SYN, ACK, RST, FIN); abnormal patterns are strong threat signals.
- Throughput: Effective data delivered per unit time; drops often indicate congestion or faults.
- Latency: Time for a packet to traverse the path; persistent spikes may reflect routing or capacity issues.
- Ports: Logical endpoints (e.g., 53/DNS, 443/HTTPS); unexpected open/active ports are common indicators.
- TTL: Time-to-Live prevents infinite routing; unusual TTL can hint at spoofing or atypical paths.
- MTU/MSS: Maximum frame/segment sizes; mismatches cause fragmentation and throughput penalties.
How to Read Typical Packet Metrics
- Count: Total number of anomalies or events in the selected window.
- Heatmap of Anomalies by Day: Calendar-style density view for spotting recurring hotspots by date/hour.
- Top Anomaly Reasons: Ranked breakdown (e.g., Port Scan, UDP Burst, Auth Failures) to prioritize response.
- Packet Volume Over Time: Time-series of packets; spikes can indicate scans, floods or workload surges.
- UDPv4 / TCP_DATAv4 / TCP_CONNECTv4: Example protocol/event classes commonly tracked; values like 468 / 211 / 50 indicate counts within the time range.