Adapting Bro into SCADA: building a specification-based intrusion detection system for the DNP3 protocol

Hui Lin, Adam J. Slagell, Catello Di Martino, Zbigniew Kalbarczyk, Ravishankar K. IyerCSIIRW 2013 · DOI

TL;DR

This paper introduces a novel intrusion detection system (IDS) for SCADA networks by adapting the Bro network analysis framework to understand the DNP3 protocol. By building a DNP3-specific parser and defining protocol validation policies, they enable deep semantic analysis of SCADA traffic, offering a more robust defense against cyberattacks than traditional IDSes.

What Problem Does This Paper Address?

Modern SCADA systems, crucial for operating critical infrastructures like power grids and water plants, are increasingly connected to public networks. This connectivity, while improving efficiency, exposes them to significant cyber threats. Traditional intrusion detection systems struggle with SCADA's proprietary network protocols, lacking the ability to understand the specific semantics of control commands and data. This limitation leaves SCADA systems vulnerable to attacks that exploit these protocol nuances, making them blind to many malicious operations. The paper aims to bridge this gap by creating an IDS that can deeply analyze DNP3 traffic, a widely used SCADA protocol.

Methods

The core of the proposed solution is the adaptation of Bro, a well-established real-time network traffic analyzer, for SCADA environments. The authors developed a DNP3-specific parser using binpac, a tool for generating protocol parsers. This parser translates raw DNP3 network packets into meaningful SCADA-specific events, extracting semantic information. These events are then fed into event handlers, where security policies are implemented as Bro scripts.

A key innovation is the "protocol validation policy." This policy uses two types of checks:

  • Intra-packet validation: Ensures that data fields within a single packet adhere to defined rules and value ranges. This helps detect malformed packets or unexpected values that could indicate attacks or system failures.
  • Inter-packet validation: Verifies that communication patterns between packets follow expected sequences (e.g., a "SELECT" command followed by an "OPERATE" command). This helps identify anomalies like replay attacks or denial-of-service attempts that disrupt normal communication flows.

Datasets

The paper describes evaluation using two types of traffic:

  • A sample traffic trace collected from a real electrical power grid in Ohio.
  • Malformed synthetic network traffic generated for testing the protocol validation policy.

No new benchmark datasets were introduced; rather, existing traffic and synthetically generated malformed packets were used to evaluate the system's robustness and detection capabilities.

Key Insights and Ideas

  • Bro's Adaptability: The paper demonstrates that Bro, a general-purpose network analysis tool, can be effectively extended to understand highly specialized industrial protocols like DNP3.
  • Semantic Analysis is Crucial: Simply looking at packet headers isn't enough for SCADA security. Understanding the meaning of DNP3 commands and data is essential for detecting sophisticated attacks.
  • Specification-Based Policies: Defining security policies based on the formal specifications of a protocol (like DNP3) provides a robust way to detect deviations from expected behavior.
  • Robustness Against Malformed Packets: The DNP3 parser, by avoiding recursive production rules, exhibits greater robustness against malformed packets compared to some existing tools like Wireshark, which can crash under such conditions.

Key Results

The DNP3 analyzer demonstrated strong performance:

  • Throughput: The DNP3 parser alone achieved a throughput of 13,950 packets per second. When the protocol validation policy was active, the throughput decreased to 9,427 packets per second, a reduction of about 30%, attributed to the intensive analysis required by the policy.
  • Robustness: The DNP3 analyzer did not crash or exhibit unreliable behavior when processing malformed packets, unlike Wireshark in one observed instance.
  • Scalability Potential: The authors estimate that the analyzer could monitor a field site with 4,500 to 9,000 devices, and that scaling up by distributing analyzers across multiple hosts could monitor even larger environments.

Limitations and Future Work

Limitations: * The current protocol validation policy is a proof of concept, focusing on basic protocol adherence. * The evaluation primarily focused on detecting malformed packets and protocol violations, not necessarily well-formatted malicious control operations.

Future Work: * Applying the DNP3 analyzer to detect well-formatted control operations from malicious users. * Developing more sophisticated correlation techniques to understand the purpose of suspicious network traffic by analyzing control operations and host activities. * Exploring the design of an intrusion prevention system based on the proposed analyzer to actively stop malicious operations.

Summary updated

BibTeX
@inproceedings{lin2013adapting,
  author = {Hui Lin and Adam J. Slagell and Catello Di Martino and Zbigniew Kalbarczyk and Ravishankar K. Iyer},
  title = {{Adapting Bro into SCADA: building a specification-based intrusion detection system for the DNP3 protocol}},
  booktitle = {CSIIRW},
  year = {2013},
  doi = {10.1145/2459976.2459982},
}
Figure