acceptodds
Withdrawn from ICLR 2027

AdaDoS: Adaptive DoS Attack via Deep Adversarial Reinforcement Learning in SDN

Withdrawn from ICLR 2027, so left out of the paper lists and the map.

Abstract

Existing defence mechanisms have demonstrated significant effectiveness in mitigating rule-based Denial-of-Service (DoS) like attacks, leveraging predefined signatures and static heuristics to identify and block malicious traffic. However, the emergence of AI-driven techniques presents new challenges to SDN security, potentially compromising the efficacy of existing defence mechanisms. In this paper, we introduce AdaDoS, an adaptive attack model that disrupt network operations while evading detection by existing DoS-based detectors through adversarial reinforcement learning (RL). Specifically, AdaDoS models the problem as a competitive game between an attacker, whose goal is to obstruct network traffic without being detected, and a detector, which aims to identify malicious traffic. AdaDoS can solve this game by dynamically adjusting its attack strategy based on feedback from the SDN and the detector. Additionally, recognising that attackers typically have less information than defenders, AdaDoS formulates the DoS-like attack as a partially observed Markov decision process (POMDP), with the attacker having access only to delay information between attacker and victim nodes. We address this challenge with a novel reciprocal learning module, where the student agent, with limited observations, enhances its performance by learning from the teacher agent, who has full observational capabilities in the SDN environment. AdaDoS represents the first application of RL to develop DoS-like attack sequences, capable of adaptively evading both machine learning-based and rule-based DoS-like attack detectors.

Primary area:
learning on graphs and other geometries & topologies
Keywords:
Network Security, DDoS

Reviews

Among:
Shape:
1234this paper, Bottom 2%
Figure 1. Mean review rating of the reviewed papers at ICLR 2027. This paper averages 1.33 over 3 reviews, above 2% of the others.
Reviewer_evZQOpenReview ↗
Rating
2
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

AdaDoS is a proposed automated attack on SDNs to cause degradation of network throughput while minimizing the likelihood of detection by an IDS. It is trained using a reciprocal policy transfer, where a teacher agent receives the true state as input, and uses it to guide the reward shaping of a student model. AdaDoS was compared to LDoS and Double-LDoS, and they found that the attack success rate was much higher against both models, with similar average bandwidth compared to Double-LDoS (meaning, both models caused similar damage to the network, but AdaDoS was less likely to be caught). They also tested the model’s robustness to noise, but found that it causes substantial decrease in attack success rate, concluding this would be a successful defense strategy against their method.

Critical strengths
  • The teacher-student reward shaping is a good idea for POMDPs, especially for automated security agents where observations can differ so broadly from the true states
  • The results on the baselines are strong both in terms of ASR and bandwidth
  • The theory is good and the proofs appear sound.
Critical weaknesses
  • Is this really “adversarial RL”? The teacher and student models are trying to align their KL divergence. They are working in conjunction, not adversarially.
  • There is no ablation of how well the student model learns without its teacher, which is a critical part of the claimed contributions of this work
  • The threat model seems unrealistic to me. Would the attacker really have observability into the SDN and IDS?
  • The reward function requires more justification. Why is the reward for congestion scaled by z-c^2?
  • The noise ranges in Section 5.4 are very wide assuming the data is normalized. Especially with the mean at 10 and std at 5, the noise is probably dominating the signal. This experiment should be re-run with less noise to determine the boundary for when the additional noise causes issues, rather than showing 3 instances where it pushes the model back to essentially random behavior
  • The brief discussion of GANs at the end seems out of place in the paper
  • The conclusion needs to be rewritten. There are numerous typos, and references undefined terms (“survival ability problem”) that appear nowhere else

Nitpicks:

  • SDN should be defined in the abstract and intro before it is used as an acronym
  • Introduction says contributions are threefold, then lists 4 contributions
  • The specific appendix with the detailed list of topologies is not mentioned
  • Table 1 needs more significant figures in some of the results; the ‘/’ in the title of Table 2 never appears
  • The two tables should also appear closer in the text to where they are referenced if possible
Questions for authors
  • How does just the student model do without the help of the teacher’s transfer learning?
  • How would the strong affect of noise shown in Section 5.4 affect deploying this model in the (noisy) real world?
  • What was the motivation for the reward function, in particular the reward for congestion?
Reviewer_5m2qOpenReview ↗
Rating
1
Confidence
3
Informativeness
2
Presentation
2
Soundness
2
Read review
Paper summary

This paper proposes AdaDoS, a reinforcement learning-based framework for adaptive DoS attacks in SDN environments. It employs a two-stage PPO policy to dynamically adjust attack timing, rate, and duration, and introduces a reciprocal teacher-student learning mechanism to address partial observability. Experiments using Mininet and Ryu across four network topologies show that AdaDoS outperforms conventional LDoS attacks, achieving attack success rates of 79.0%–87.1%. The authors also evaluate the effectiveness of reciprocal learning and robustness to observation noise.

Critical strengths

1.The paper addresses the important problem of adaptive DoS attacks in SDN environments. Considering limited attacker observability makes the threat model more practically relevant and provides useful insights into potential vulnerabilities of existing detection mechanisms. 2.The authors evaluate AdaDoS across four network topologies and demonstrate improvements over conventional LDoS attacks in terms of attack success rate and bandwidth disruption. These results provide preliminary evidence of the effectiveness of the proposed approach.

Critical weaknesses

1.Insufficient literature review and unclear novelty. The related work is not sufficiently comprehensive, particularly regarding recent studies on adaptive DoS/DDoS attacks and reinforcement learning-based network security. Most of the discussed studies were published in or before 2023, while relevant developments from 2024 onward receive limited attention. Given the extensive existing literature, the paper does not clearly establish its novelty or sufficiently differentiate the proposed method from existing approaches. 2.Insufficient experimental validation. The evaluation primarily compares AdaDoS against conventional LDoS attacks with different parameters, without sufficiently considering stronger adaptive or reinforcement learning-based baselines. Moreover, the lack of systematic ablation studies makes it difficult to assess the individual contributions of the two-stage decision mechanism and reciprocal learning. These limitations weaken the experimental support for the claimed methodological improvements.

Questions for authors

1.The literature review is insufficient, particularly regarding recent studies (2024–2026) on adaptive DoS/DDoS attacks and reinforcement learning-based approaches. This also leads to inadequate comparisons with existing methods, making it difficult to assess the novelty and effectiveness of the proposed approach. 2.Why does the evaluation primarily use conventional LDoS baselines rather than stronger adaptive or reinforcement learning-based approaches? 3.What are the individual contributions of the two-stage decision mechanism and reciprocal learning? Can the authors provide corresponding ablation results? 4.How does the deployed attacker obtain the detector feedback and congestion information required for online policy updates? 5.Figure 6 shows a substantial performance degradation under delay noise. How robust is AdaDoS under realistic network fluctuations? 6.How does the convergence result in Theorem 2 apply to the actual PPO-based reciprocal learning algorithm? 7.Why are attack costs defined using different aggregation intervals for AdaDoS and LDoS in Appendix E.4?

Reviewer_BMFAOpenReview ↗
Rating
1
Confidence
3
Informativeness
1
Presentation
1
Soundness
1
Read review
Paper summary

This paper introduces the AdaDoS, an adaptive attack model that disrupts network operations while evading detection by existing DoS-based detectors through adversarial reinforcement learning (RL). AdaDoS combines a two-stage RL policy with teacher–student reciprocal learning to generate adaptive DoS attacks under partial observation. The application is relevant, but I am not convinced that the paper establishes a sufficiently clear technical contribution or novelty. Meanwhile, the presentation is quite poor, and the theoretical justification shows a weak connection to its experiments.

Critical strengths
  1. The paper addresses a relevant security problem for adaptive DoS attacks.

  2. The evaluation considers multiple network topologies and detector configurations, with additional experiments on observation noise.

Critical weaknesses
  1. The novelty and the technical contribution of this work are limited and unclear. The method combines existing components, including PPO, reward shaping, and teacher–student policy transfer. The paper repeatedly describes its components as novel, but does not clearly explain what technical advances they provide over existing approaches.

  2. The theory in this paper shows a weak connection with its actual methods. Theorem 1 bounds the return difference when teacher and student policies are already close, but does not establish that the proposed reciprocal updates achieve this condition.

  3. The overall writing of this paper is poor; the structure is chaotic, and this paper is clearly not well polished.

Questions for authors

See the weaknesses.

From OpenReview, CC BY 4.0. Authors as signed there; shown unedited.

Related papers

Discussion (0)

Sign in to comment.