Indian Institute of Technology
Department of Computer Science & Engineering
Research Intern (Network Security) - 2023
Understanding the Threat
Slowloris attacks silently drain server resources by keeping connections open, often bypassing traditional DDoS defenses.
Their subtle nature makes them hard to detect but highly disruptive.
Incomplete requests consume resources, blocking legitimate users
Layered Defense Strategy Funnel
Built a multi-tiered defense to counter Slowloris attacks by layering Nginx buffering, Apache timeout tuning, packet-level filtering, and dynamic IP banning. Each layer filters traffic independently, creating a narrowing funnel that blocks malicious connections while preserving legitimate ones, ensuring stable server performance under sustained, low-bandwidth attacks.
.png)
.png)
-
Target Environment: Legacy Linux servers without cloud-native WAFs or CDN-level protection.
-
Problem Defined: Application-layer DoS via slow, persistent HTTP headers escaping traditional bandwidth-based detection.
-
Key Requirements:
-
Low overhead, self-contained mitigation stack
-
Compatibility with Apache/Nginx stack
-
Configurable timeout + connection thresholds
-
-
Engineering Decisions:
-
Chose Nginx as reverse proxy for early connection buffering
-
Hardened Apache with mod_reqtimeout for idle-session termination
-
Integrated Fail2Ban for automated IP-level banning
-
Used Wireshark logs for attack pattern validation
-
Ran A/B tests on timeout + threshold configs to optimize impact
-
-
Scope Constraints: No access to CDN/WAF or kernel-level firewall customization.
PRD Highlights
-
Maintained ≥99% server uptime during sustained simulated Slowloris attacks
-
Dropped over 80% of idle open connections within defined timeout thresholds
-
Kept CPU overhead under 10%, ensuring stability on low-resource VMs
-
Achieved ≤5% false positives, maintaining access for legitimate users
-
Successfully integrated mitigation stack with no external tools, built entirely on open-source solutions
-
Delivered a replicable, modular defense strategy tailored for legacy Apache infrastructure
Project Outcomes
© 2025 Aviraj Dongare. All rights reserved.