An official website of the United States government
Phishing attacks remain one of the most prevalent and pervasive cybersecurity concerns. Voice phishing (i.e., vishing) is an emerging type of phishing attack where malicious actors use audio channels to steal sensitive information from victims. However, vishing detection is a challenging task due to its real-time nature and the limited availability of datasets. To help address the concern of vishing detection, this study proposes the vishing generative pretrained transformer (VishGPT). VishGPT adopts the computational design paradigm and incorporates novel reinforcement learning-based large language model fine-tuning and synthetic data model pretraining to automatically detect vishing attempts in real time. We evaluated VishGPT using a series of benchmark experiments, where we empirically demonstrated its improvement over state-of-the-art vishing detection and audio classification models. The results suggest that our proposed VishGPT achieved state-of-the-art performance in terms of accuracy (86.18%), precision (90.63%), recall (85.02%), and F1-score (87.74%). VishGPT offers practical value to cybersecurity professionals, end users, and academia. Additionally, VishGPT provides important design principles in the form of a custom proximal policy optimization (PPO) reward function and synthetic pretraining to the information systems knowledge base.
more »
« less
This content will become publicly available on June 1, 2027
AUTOMATICALLY DETECTING VOICE PHISHING: A LARGE AUDIO MODEL APPROACH
Phishing attacks remain one of the most prevalent and pervasive cybersecurity concerns. Voice phishing (i.e., vishing) is an emerging type of phishing attack where malicious actors use audio channels to steal sensitive information from victims. However, vishing detection is a challenging task due to its real-time nature and the limited availability of datasets. To help address the concern of vishing detection, this study proposes the vishing generative pretrained transformer (VishGPT). VishGPT adopts the computational design paradigm and incorporates novel reinforcement learning-based large language model fine-tuning and synthetic data model pretraining to automatically detect vishing attempts in real time. We evaluated VishGPT using a series of benchmark experiments, where we empirically demonstrated its improvement over state-of-the-art vishing detection and audio classification models. The results suggest that our proposed VishGPT achieved state-of-the-art performance in terms of accuracy (86.18%), precision (90.63%), recall (85.02%), and F1-score (87.74%). VishGPT offers practical value to cybersecurity professionals, end users, and academia. Additionally, VishGPT provides important design principles in the form of a custom proximal policy optimization (PPO) reward function and synthetic pretraining to the information systems knowledge base.
more »
« less
- PAR ID:
- 10684025
- Publisher / Repository:
- MIS Quarterly
- Date Published:
- Journal Name:
- MIS Quarterly
- ISSN:
- 2162-9730
- Subject(s) / Keyword(s):
- Deep learning voice phishing computational design science information systems large language models reinforcement learning audio classification
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
null (Ed.)Objective To understand how aspects of vishing calls (phishing phone calls) influence perceived visher honesty. Background Little is understood about how targeted individuals behave during vishing attacks. According to truth-default theory, people assume others are being honest until something triggers their suspicion. We investigated whether that was true during vishing attacks. Methods Twenty-four participants read written descriptions of eight real-world vishing calls. Half included highly sensitive requests; the remainder included seemingly innocuous requests. Participants rated visher honesty at multiple points during conversations. Results Participants initially perceived vishers to be honest. Honesty ratings decreased before requests occurred. Honesty ratings decreased further in response to highly sensitive requests, but not seemingly innocuous requests. Honesty ratings recovered somewhat, but only after highly sensitive requests. Conclusions The present results revealed five important insights: (1) people begin vishing conversations in the truth-default state, (2) certain aspects of vishing conversations serve as triggers, (3) other aspects of vishing conversations do not serve as triggers, (4) in certain situations, people’s perceptions of visher honesty improve, and, more generally, (5) truth-default theory may be a useful tool for understanding how targeted individuals behave during vishing attacks. Application Those developing systems that help users deal with suspected vishing attacks or penetration testing plans should consider (1) targeted individuals’ truth-bias, (2) the influence of visher demeanor on the likelihood of deception detection, (3) the influence of fabricated situations surrounding vishing requests on the likelihood of deception detection, and (4) targeted individuals’ lack of concern about seemingly innocuous requests.more » « less
-
This study examines how artificial intelligence (AI) can help with voice phishing (vishing) attacks, with a particular emphasis on deepfake technologies and AI-driven voice synthesis. It examines the strategies used by cybercriminals, assesses the effectiveness of the present defenses, and identifies difficulties in identifying and preventing such attacks. The results show that to combat the increasing complexity of vishing strategies, there is an urgent need for sophisticated detection systems and preventive actions. Future directions include the creation of cooperative policy frameworks to control the misuse of AI and easily accessible solutions for small enterprises.more » « less
-
This study examines how artificial intelligence (AI) can help with voice phishing (vishing) attacks, with a particular emphasis on deepfake technologies and AI-driven voice synthesis. It examines the strategies used by cybercriminals, assesses the effectiveness of the present defenses, and identifies difficulties in identifying and preventing such attacks. The results show that to combat the increasing complexity of vishing strategies, there is an urgent need for sophisticated detection systems and preventive actions. Future directions include the creation of cooperative policy frameworks to control the misuse of AI and easily accessible solutions for small enterprises.more » « less
-
Phishing websites remain a persistent security threat. Thus far, machine learning approaches appear to have the best potential as defenses. But, there are two main concerns with existing machine learning approaches for phishing detection. The first is the large number of training features used and the lack of validating arguments for these feature choices. The second concern is the type of datasets used in the literature that are inadvertently biased with respect to the features based on the website URL or content. To address these concerns, we put forward the intuition that the domain name of phishing websites is the tell-tale sign of phishing and holds the key to successful phishing detection. Accordingly, we design features that model the relationships, visual as well as statistical, of the domain name to the key elements of a phishing website, which are used to snare the end-users. The main value of our feature design is that, to bypass detection, an attacker will find it very difficult to tamper with the visual content of the phishing website without arousing the suspicion of the end user. Our feature set ensures that there is minimal or no bias with respect to a dataset. Our learning model trains with only seven features and achieves a true positive rate of 98% and a classification accuracy of 97%, on sample dataset. Compared to the state-of-the-art work, our per data instance classification is 4 times faster for legitimate websites and 10 times faster for phishing websites. Importantly, we demonstrate the shortcomings of using features based on URLs as they are likely to be biased towards specific datasets. We show the robustness of our learning algorithm by testing on unknown live phishing URLs and achieve a high detection accuracy of 99.7%.more » « less
