Patrick Traynor’s research while affiliated with University of Florida and other places

... Only 17 (22%) papers released their source code, and 5 papers (6%) release the exact prompts used to realize the attack. In contrast, 57 papers (72%) do not provide such low-level details (a result which echoes [62,161]). A detailed explanation of this analysis is provided in Appendix A-D (and these results are also shown in Table IV). ...

Reference:

SoK: On the Offensive Potential of AI

... Peeters et al. [30] designed an intelligent approach named LeopardSeal to identify the call eavesdropping through Audio Rogue Base Station (ARBS). The developed Leop-ardSeal approach identified the attacks effectively. ...

Reference:

An ensemble learning approach for intrusion detection in IoT-based smart cities

... By adding high frequency components out of the voice band, attackers can launch spectrum addition attacks to generate the audio that can be interpreted by machines but incomprehensible to humans [1]. In contrast, by removing the frequency components of weak strength from the audio spectrum, an attacker can launch spectrum reduction attacks to generate the audio that can be perceived by humans but cannot be correctly interpreted by machines [3], [2]. Moreover, attackers can manipulate the spectrum magnitude with a specific filter to bypass the spectrum-based detection mechanisms [69]. ...

Reference:

Compensating Removed Frequency Components: Thwarting Voice Spectrum Reduction Attacks

... The various attacks that can occur in the context of SMS OTP include replay attacks, brute force attacks, poor OTP creation attacks, man-in-the-middle attacks, phishing attacks, redirection attacks, and malware attacks. Due to the high risk of attack, the National Institute of Standards and Technology (NIST) [2] and the National Cyber Security Center (NCSC) [3] have removed SMS as a recommended OTP delivery channel. NIST states that the use of SMS is vulnerable to social engineering and malware attacks through endpoint compromise [2]. ...

Reference:

Device risk analysis protocol for SMS-based OTP Authentication

... However, it is also constrained by human assumptions, making it unsuitable for complex devices like Ethernet and USB DMA. FirmWire [29] offers human-crafted peripheral models for baseband firmware. ...

Reference:

IoT Firmware Emulation and Its Security Application in Fuzzing: A Critical Revisit

... Research has shown that DNNs are also vulnerable to adversarial attacks (Abdullah et al., 2021b;Finlayson et al., 2019;Ma et al., 2021;Szegedy, 2013;Wang et al., 2022Wang et al., , 2023. As modern ASR systems increasingly incorporate Semantic Web-driven technologies to better understand speech, the potential for adversarial attacks becomes even more complex. ...

Reference:

Semantic Web-Driven Targeted Adversarial Attack on Black Box Automatic Speech Recognition Systems

... Abdullah et al. [171] contribute to the growing body of work on adversarial attacks by classifying and evaluating the various vulnerabilities of ASR systems, particularly the endto-end architectures that are commonly used. Their research shows that adversarial attacks generally do not transfer well between different ASR models, calling for more adaptive and comprehensive defense strategies that can account for the unique challenges posed by these architectures. ...

Reference:

Beyond Vulnerabilities: A Comprehensive Survey of Adversarial Attacks Across Domains

... According to industry research, merchants are projected to incur losses of USD 130 billion from fraudulent transactions between 2018 and 2023 [1]. Many financial institutions allocate a security budget ranging from 20% to 30%, known as extended detection and response (XDR), considered a top priority in their security investments [2]. ...

Reference:

Efficiency of Federated Learning and Blockchain in Preserving Privacy and Enhancing the Performance of Credit Card Fraud Detection (CCFD) Systems

... Adversarial attacks [3][4][5][6], for instance, aim to confuse the SV system by introducing well-crafted perturbations into the speech signals. In recent years, many studies [7][8][9][10][11][12] have conducted effective adversarial attacks on SV systems. On the other hand, spoofing attacks [13,14] attempt to mimic the timbre of target speakers through methods commonly including impersonation, replay, voice conversion, and speech synthesis. ...

Reference:

Speaker Contrastive Learning for Source Speaker Tracing

... The consequences become dangerous when these attacks are directed towards vital infrastructure. They can potentially disrupt essential services such as healthcare, transportation, and emergency response systems, as noted in various studies [25,26]. It is worth highlighting that the increasing integration of Internet of Things (IoT) devices into critical infrastructure renders them particularly susceptible to exploitation. ...

Reference:

Automated Network Incident Identification through Genetic Algorithm-Driven Feature Selection