Adversarial Sampling Attacks Against Phishing Detection

Abstract

Phishing websites trick users into believing that they are interacting with a legitimate website, and thereby, capture sensitive information, such as user names, passwords, credit card numbers and other personal information. Machine learning appears to be a promising technique for distinguishing between phishing websites and legitimate ones. However, machine learning approaches are susceptible to adversarial learning techniques, which attempt to degrade the accuracy of a trained classifier model. In this work, we investigate the robustness of machine learning based phishing detection in the face of adversarial learning techniques. We propose a simple but effective approach to simulate attacks by generating adversarial samples through direct feature manipulation. We assume that the attacker has limited knowledge of the features, the learning models, and the datasets used for training. We conducted experiments on four publicly available datasets on the Internet. Our experiments reveal that the phishing detection mechanisms are vulnerable to adversarial learning techniques. Specifically, the identification rate for phishing websites dropped to 70% by manipulating a single feature. When four features were manipulated, the identification rate dropped to zero percent. This result means that, any phishing sample, which would have been detected correctly by a classifier model, can bypass the classifier by changing at most four feature values; a simple effort for an attacker for such a big reward. We define the concept of vulnerability level for each dataset that measures the number of features that can be manipulated and the cost for each manipulation. Such a metric will allow us to compare between multiple defense models.

Ontology-based Attack Graph Enrichment

“…In this approach, initial network configuration and description of exploits serve as input for the minimal attack graph generation. Shirazi et al [18] present an approach for modeling attack-graph generation and analysis problems as a planning problem. They present a tool called AGBuilder that generates attack graphs using the Planning Domain Definition Language (PDDL) from extracted vulnerability information.…”

Feature Importance Guided Attack: A Model Agnostic Adversarial Attack

“…Recently a number of authors have proposed and developed adversarial evasion attacks against phishing detection models. Shirazi et al propose an adversarial random sampling attack [46]. They use publicly available datasets where every feature is binned to either -1, 0 or 1; where -1 is a legitimate instance, 0 is suspicious and 1 indicates phishing.…”

Bypassing Detection of URL-based Phishing Attacks Using Generative Adversarial Deep Neural Networks

“…We believe that proposing new techniques to synthesize phishing attempts is quite significant in creating a defensive mechanism that can prevent zero-day phishing attempts. While there are approaches that apply GAN for generating adversarial phishing examples [14], to the best of our knowledge, this is the first work that synthesizes URL adversarial phishing examples in order to evade sophisticated phishing detection techniques, which rely on semantic relationships between the components of URLs. Vector Machines, Naïve Bayes, and k-Nearest Neighbor.…”

“…Adversarial models have been recently used to evade machine learning classifiers. GAN has been used in intrusion detection [12], malware detection [9], and spam detection [31], and phishing [14].…”