Abstract : In recent years, we have seen the rise of application specific attacks that exploit the vulnerabilities in the network
protocols (HTTP, DNS, SMTP, other) and try to overwhelm the server application, not just the connectivity pipe.In this
paper, we propose an advanced DoS Threat Analytics System (DTAS) to mitigate the full range of DoS network attacks –
not just volumetric, based on comprehensive collaborative detection algorithms, implemented in the Elasticsearch Big Data
platform. DTAS security solution is driven by powerful threat detection algorithms that: a) dissects all attack probabilities
in the network traffic, b) Uses behavioral analytics to correlate multiple parameters and generate multi-vector
representations, c) Employs dynamic challenges to verify normal versus attack traffic. The DTAS analytics engine
analyzes multiple IP attributes within TCP and UDP flows, ICMP, HTTP and DNS traffic, count, frequency, headers,
payloads, detecting covert traffic, amplification attacks trying to target the services on the network. By measuring all these
attributes, our system creates a multi-vector heuristic representation of the normal or baseline traffic flows. We have used
datasets from UCLA, downloaded traces from real world incidents and tested the efficacy of the system with various largescale simulated DoS attacks in the test network. Our experiments show that the DTAS framework can detect DoS attacks
in real time, without impacting the latency to benign traffic in the network and with accuracy up to 95% detection rate for
attacks.