Performance Evaluation of DCA and SRC on a Single Bot Detection

Yousof Al-Hammadi, Uwe Aickelin, Julie Greensmith

Abstract

Malicious users try to compromise systems using new techniques. One of the recent techniques used by the attacker is to perform complex distributed attacks such as denial of service and to obtain sensitive data such as password information. These compromised machines are said to be infected with malicious software termed a "bot". In this paper, we investigate the correlation of behavioural attributes such as keylogging and packet flooding behaviour to detect the existence of a single bot on a compromised machine by applying (1) Spearman's rank correlation (SRC) algorithm and (2) the Dendritic Cell Algorithm (DCA). We also compare the output results generated from these two methods to the detection of a single bot. The results show that the DCA has a better performance in detecting malicious activities.

Introduction

Traditional botnet detection often focuses on network-level analysis, identifying patterns across massive amounts of traffic. However, modern attackers frequently evade these defenses through encryption or non-standard ports. This research shifts the focus to "host-based" detection—identifying an infected machine by its behavior. The primary challenge is that individual bot actions, such as recording keystrokes or sending packets, can mimic legitimate software. To address this, we evaluate two correlation strategies to identify bots by how they subvert a system once inside.

Methodology

The study compares two distinct approaches for correlating behavioral attributes:

  1. Spearman’s Rank Correlation (SRC): A standard statistical measure that evaluates the relationship between variables, such as keylogging activity and network response times.
  2. Dendritic Cell Algorithm (DCA): An Artificial Immune System (AIS) algorithm inspired by human dendritic cells. It processes three signal types: PAMPs (strong evidence of abnormality), Danger Signals (indicators of system stress), and Safe Signals (indicators of normal behavior).

To improve detection accuracy and reduce false alarms in low-data environments, the authors introduce the MAC (MCAV Antigen Coefficient) metric, which refines standard DCA outputs.

Experimental Setup

Controlled experiments were conducted using:

  • Malicious Software: Spybot and Sdbot, which feature keylogging and packet-flooding (DDoS) capabilities.
  • Benign Software: Legitimate IRC clients (IceChat), Wordpad, and standard command-line tools.
  • Scenarios: Five test cases ranging from inactive bots to combined keylogging and flooding attacks.

Results and Discussion

The DCA consistently outperformed the SRC. While the SRC struggled to distinguish between a bot's high-speed packet sending and legitimate activity without a secondary trigger, the DCA’s multi-signal fusion provided superior discrimination.

Bot's MAC values across different experimental scenarios.

A critical advantage of the DCA is its use of "Safe Signals," which suppress false alarms during temporary spikes in system stress. Statistical analysis using the Mann-Whitney test confirmed that the difference in DCA scores between bots and normal software was highly significant (p < 0.001). As shown in the comparison below, the legitimate IRC client maintained near-zero MAC values across the same scenarios where the bot was clearly identified.

Legitimate IRC client's MAC values across the same experimental scenarios, showing minimal false positives.

Conclusion

The Dendritic Cell Algorithm provides a robust framework for host-based bot detection by mimicking the human immune system's data fusion capabilities. While both algorithms failed to detect "idle" bots that execute no malicious payload, the DCA demonstrated high accuracy in active attack scenarios. Future work will focus on automating signal mapping and applying the DCA to decentralized Peer-to-Peer (P2P) botnets.

Summary updated

Abstract

Malicious users try to compromise systems using new techniques. One of the recent techniques used by the attacker is to perform complex distributed attacks such as denial of service and to obtain sensitive data such as password information. These compromised machines are said to be infected with malicious software termed a "bot". In this paper, we investigate the correlation of behavioural attributes such as keylogging and packet flooding behaviour to detect the existence of a single bot on a compromised machine by applying (1) Spearman's rank correlation (SRC) algorithm and (2) the Dendritic Cell Algorithm (DCA). We also compare the output results generated from these two methods to the detection of a single bot. The results show that the DCA has a better performance in detecting malicious activities.
BibTeX
@inproceedings{alhammadi2009performance,
  author = {Yousof Al-Hammadi and Uwe Aickelin and Julie Greensmith},
  title = {{Performance Evaluation of DCA and SRC on a Single Bot Detection}},
  booktitle = {Nottingham ePrints (University of Nottingham)},
  year = {2009},
}

Loading related content...

Figure