Muhammad Sajidur Rahman

Muhammad Sajidur Rahman, Blas Kojusner, Ryon Kennedy et al.

Frequently advised secure development recommendations often fall short in practice for app developers. Tool-driven (e.g., using static analysis tools) approaches lack context and domain-specific requirements of an app being tested. App developers struggle to find an actionable and prioritized list of vulnerabilities from a laundry list of security warnings reported by static analysis tools. Process-driven (e.g., applying threat modeling methods) approaches require substantial resources (e.g., security testing team, budget) and security expertise, which small to medium-scale app dev teams could barely afford. To help app developers securing their apps, we propose SO{U}RCERER, a guiding framework for Android app developers for security testing. SO{U}RCERER guides developers to identify domain-specific assets of an app, detect and prioritize vulnerabilities, and mitigate those vulnerabilities based on secure development guidelines. We evaluated SO{U}RCERER with a case study on analyzing and testing 36 Android mobile money apps. We found that by following activities guided by SO{U}RCERER, an app developer could get a concise and actionable list of vulnerabilities (24-61% fewer security warnings produced by SO{U}RCERER than a standalone static analyzer), directly affecting a mobile money app's critical assets, and devise a mitigation plan. Our findings from this preliminary study indicate a viable approach to Android app security testing without being overwhelmingly complex for app developers.

Hadi Abdullah, Muhammad Sajidur Rahman, Christian Peeters et al.

Automatic Speech Recognition (ASR) systems convert speech into text and can be placed into two broad categories: traditional and fully end-to-end. Both types have been shown to be vulnerable to adversarial audio examples that sound benign to the human ear but force the ASR to produce malicious transcriptions. Of these attacks, only the "psychoacoustic" attacks can create examples with relatively imperceptible perturbations, as they leverage the knowledge of the human auditory system. Unfortunately, existing psychoacoustic attacks can only be applied against traditional models, and are obsolete against the newer, fully end-to-end ASRs. In this paper, we propose an equalization-based psychoacoustic attack that can exploit both traditional and fully end-to-end ASRs. We successfully demonstrate our attack against real-world ASRs that include DeepSpeech and Wav2Letter. Moreover, we employ a user study to verify that our method creates low audible distortion. Specifically, 80 of the 100 participants voted in favor of all our attack audio samples as less noisier than the existing state-of-the-art attack. Through this, we demonstrate both types of existing ASR pipelines can be exploited with minimum degradation to attack audio quality.

Hadi Abdullah, Muhammad Sajidur Rahman, Washington Garcia et al.

Automatic speech recognition and voice identification systems are being deployed in a wide array of applications, from providing control mechanisms to devices lacking traditional interfaces, to the automatic transcription of conversations and authentication of users. Many of these applications have significant security and privacy considerations. We develop attacks that force mistranscription and misidentification in state of the art systems, with minimal impact on human comprehension. Processing pipelines for modern systems are comprised of signal preprocessing and feature extraction steps, whose output is fed to a machine-learned model. Prior work has focused on the models, using white-box knowledge to tailor model-specific attacks. We focus on the pipeline stages before the models, which (unlike the models) are quite similar across systems. As such, our attacks are black-box and transferable, and demonstrably achieve mistranscription and misidentification rates as high as 100% by modifying only a few frames of audio. We perform a study via Amazon Mechanical Turk demonstrating that there is no statistically significant difference between human perception of regular and perturbed audio. Our findings suggest that models may learn aspects of speech that are generally not perceived by human subjects, but that are crucial for model accuracy. We also find that certain English language phonemes (in particular, vowels) are significantly more susceptible to our attack. We show that the attacks are effective when mounted over cellular networks, where signals are subject to degradation due to transcoding, jitter, and packet loss.