Enhancing Adversarial Defense in Phishing Detection Using Deep Reinforcement Learning

Adversarial Robustness of Phishing Email Detection: A Comparative Study of TF-IDF + Logistic Regression and Fine-Tuned DistilBERT

Phishing emails remain one of the most persistent cybersecurity threats, and machine-learning classifiers are widely used to detect them. Most reported detection accuracies, however, are measured on clean, in-distribution test data rather than on emails deliberately altered to evade detection. This paper reports a controlled, pairwise comparison of two phishing-detection approaches a TF-IDF + Logistic Regression baseline and a fine-tuned DistilBERT transformer trained on a unified corpus of 82,255 emails drawn from six public datasets and evaluated under three conditions: normal in-distribution, synthetic phishing, and adversarial phishing. Both models exceeded 98% accuracy on clean data yet degraded sharply under adversarial testing: TF-IDF + LR fell to 64.00% (a 34.59-percentage-point drop) and DistilBERT fell to 63.64% (a 35.40-percentage-point drop) a gap of only 0.36 percentage points, equivalent to a single email in the 275-sample adversarial test set. LIME, SHAP, and attention-rollout analysis indicate the two models relied on different evidence yet showed similar vulnerability. Pairwise error analysis shows the models agreed on 54.9% of adversarial samples but each made a similar number of exclusive errors (24 and 25 respectively), indicating partly complementary rather than identical failure modes. The results show that clean-data accuracy does not predict adversarial robustness, and that adversarial testing should be a standard part of phishing-detection evaluation.

Detecting adversarial evasion in deep learning intrusion detection systems using explainable AI

Deep learning based network intrusion detection systems (IDS) can achieve strong traffic classification performance, but their resilience to adversarial manipulation remains a critical concern. This study evaluates the adversarial robustness of Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models in a multiclass intrusion detection setting using the Train_Test_Network dataset with ten traffic classes. The models were trained on true sliding flow-window sequences under a unified preprocessing pipeline to support fair comparison. Adversarial robustness was first assessed under a white-box Fast Gradient Sign Method (FGSM) setting and then broadened through additional FGSM and Projected Gradient Descent (PGD) stress testing. SHapley Additive exPlanations (SHAP) were further used to analyse explanation instability under clean and adversarial conditions, and explanation-drift features were evaluated as a secondary adversarial detection signal. Under clean evaluation, both models achieved strong and nearly identical performance, with accuracies of 0.9614 for LSTM and 0.9615 for GRU and weighted F1-scores of 0.9597 and 0.9598, respectively. Under the main FGSM condition, performance declined substantially: the LSTM achieved adversarial accuracy of 0.6094 and weighted F1-score of 0.6290 with an evasion rate of 37.38%, while the GRU achieved adversarial accuracy of 0.5130 and weighted F1-score of 0.5690 with an evasion rate of 47.02%. The broader robustness sweep showed that iterative PGD exposed stronger fragility than FGSM alone. SHAP analysis indicated that adversarial perturbation altered both prediction outcomes and local explanation structure. A learned explanation-driven detector improved over the rule-based baseline, while larger-scale validation confirmed that explanation drift remained informative, though not perfectly separable, at broader scale. Overall, the results show that strong clean performance does not imply adversarial robustness, and that explanation drift provides a useful auxiliary signal for adversarial monitoring in recurrent IDS models.

Enhancing robustness of deep learning-based malware detection against adversarial attacks

Adversarial Machine Learning in Cybersecurity: A Survey of Attacks, Defenses, and Robustness Evaluation

The rise of artificial intelligence (AI) and machine learning (ML) in cybersecurity has made Adversarial Machine Learning (AML) a key field of research. Though ML-based systems are more effective in intrusion detection, malware analysis, spam filtering and authentication, they are also susceptible to adversarial attacks that modify input samples, modify ML models or alter training data. The review explores the key adversarial attack classes: poisoning, evasion, model extraction, model inversion, and membership inference and also white-box, black-box, and grey-box threat models. It also provides an introduction to important defence methods like adversarial training, feature squeezing, defensive distillation, robust optimisation, detection-based methods, and ensemble learning. In addition, robustness evaluation metrics, benchmark datasets and attack assessment techniques to measure model robustness are highlighted. Lastly, the emerging trends are discussed in order to find future research directions in the field of creating trustworthy and resilient AI-based cybersecurity systems, such as Explainable AI, Federated Learning, Large Language Models, Autonomous Cyber defence, and Regulatory structures.

Adversarial Deepfake Generation and an Investigation of Purification-Based Adversarial Detection

This paper describes the participation of team"Go To Germany"in the ImageCLEF 2026 Deepfake Detection and Generation Task. For the image generation task, we employ FLUX.1-dev with PuLID for identity-preserving face synthesis, combined with a multi-model PGD adversarial attack targeting 12 detectors simultaneously (DiffJPEG-in-loop, MI/DI/EoT, adaptive weighting, two-stage warm-start). Our approach achieved 90% evasion against organizer detectors and 57.6% against participant detectors, with a final generation score of 0.4170. For the image detection task, we combine two complementary detectors - SigLIP+DINOv2 for AI-generated images and GenD-DINOv3 for face manipulations - in a max-probability ensemble, achieving 99.4% accuracy on baseline deepfakes but suffering from high false-positive rates on real images, resulting in a final detection score of 0.6986. Beyond the official submission, we conducted a self-initiated investigation of purification-based adversarial detection, comparing three families of detection signals across six detectors that share a CLIP ViT-L/14 backbone. We find that raw $|\Delta \text{logit}|$ under median-3 purification, applied through the EFFORT detector, separates adversarial inputs from clean inputs with AUROC 0.81-0.98 across four adversarial source types - a finding that refutes the simple backbone-preservation hypothesis and exposes a sharp JPEG-quality cliff at Q70 where the signal collapses.

Adversarial Malware Can Be Both Evasive and Deceiving: a Gradient-based Attack Against Prediction and Explainability in Windows PE Malware Detection

In the last years, several gradient-based attacks have been developed to disclose vulnerabilities of deep neural models and strengthen evasion and transferability abilities of adversarial examples created in various domains. In Windows PE malware detection, various gradient-based methods have been recently tested to optimize the editing or injection of adversarial bytes in unused file sections, and create adversarial malicious PE files to evade deep neural models developed for malware detection. On the other hand, gradient information is commonly used in eXplainable AI (XAI) to obtain explanations for opaque decision-making processes of deep neural black-boxes, while Adversarial XAI has recently emerged as an Adversarial Learning field to investigate the security landscape of XAI. In this study, we describe a holistic, gradient-based attack method, named GAME4EXE, formulated to optimize editing of adversarial bytes in the DOS Header and DOS Stub sections of Windows PE malware. The optimization is driven by the dual adversary purpose of generating adversarial malware that evades a target deep neural model, and aligning the explanation of the evading decisions with that of a goodware-like explanation. A preliminary evaluation shows that a gradient-based attack can be effectively formulated in the Windows PE malware domain, to equip the evasion of a deep neural model with the capability to coherently deceive its decision process explainer and harden attack detection.