Grant Hernandez's research while affiliated with FIU Herbert Wertheim College of Medicine and other places

... Cao et al. [6], Johnson et al. [29], and Zhou [62], all leverage symbolic execution to learn satisfying values to bypass peripheral checks. Hernandez et al. [27] achieve full-system emulation of closed-source Shannon baseband firmware by adding missing architectural and peripheral support in QEMU, they later demonstrate that such an approach can be extended to other basebands [26]. In contrast to the aforementioned approaches, Milburn et al. [38] build a custom emulator and peripheral models to rehost an automotive instrument cluster; they use their emulator to aid in reverse-engineering the firmware's UDS commands. ...

... seccomp filters use BPF programs to decide, based on the system call number and arguments, whether a given call is allowed or not. A more recent application of eBPF is LBM [62], which protects the Linux kernel from malicious peripherals such as USB, Bluetooth, and NFC. LBM places interposition hooks, through the implementation of new eBPF program types, right beneath a peripheral's protocol stack and above the peripheral's controller driver, so that it can guarantee that eBPF programs can filter all inputs from the device and all outputs from the host. ...

... In this work, researchers built a card skimmer detector that can be used at physical payment terminals such as ATMs and gas stations. In another work, Scaife et al. [46] did a survey of gas pump card skimmer detection techniques including Bluetooth skimmer detection on iOS and Android apps, to identify common skimmer detection characteristics. ...

... However, unlike in these works, we consider systems in which samples of correct usage of the functions and features may not be available. Our use of symbolic constraints in learning differs from the approach in [27] as our constraints are not on the input variables and, instead, involve internal state of the computation, which is important for detecting unwanted feature interactions. ...

... However, the pressure of time to market of IoT development increasingly raises security and privacy concerns [61]. Firmware of IoT devices could contain vulnerabilities, which have already caused destructive attacks [42]. IoT firmware security analysis is considered an effective approach to ensuring the security of IoT devices [26]. ...

... Furthermore, the TEE-based SMC protocol was proved by simulation, and some benchmarks compared to ABY [42] are presented to show its efficiency. To alleviate the burden of trust on the enclave, Choi et al. [30] explored a balanced method to split a function . Instead of evaluating the entire function in the enclave, the proposed 2P-SFE protocol allows the designer to choose which components should be evaluated within the enclave. ...

... Testing baseband firmware using fuzzing was shown in [66], [28] which focuses on GSM, but only targets SMS and SMS cell broadcast. Previous approaches to scaling up cellular testing include developing a wireless testbed and incorporating feedback from multiple UEs to detect faults [17], [27]. Most previous work performs over-the-air testing, which is difficult to fully scale and relies on manufacturer logs from UEs, which may not be detailed enough or available at all, to triage faults. ...

... Based on trusted computing technology, a virtual trusted root is configured for each container [9]. Although a certain degree of trustworthiness assessment can be achieved, container instances in a single server in a container cloud are typically in the tens [15]. Thus, a significant performance loss is caused. ...

... Research has shown that current defense measures, such as virus scanners, are not effective to withstand emerging smartphone security threats, simply because they are signature-based, so they can only detect known malware (e.g., worms, viruses, and Trojan horses)( (Chen & etal.. 2017), (Hernandez and Butler. 2018), (Reshetova, Bonazzi, and Asokan. 2017), . Malware writers are writing malware that can easily bypass signature-based virus scanners and enetrate the operating system to execute malicious code. New and unknown malware presents serious security threats on smartphones; therefore, there is a critical need to design robust security counterm ...

... We first conduct an analysis of Linux Audit's adherence to real-time scheduling principles, discovering that while Audit introduces overheads and increased variance to each syscall, it does not introduce inter-application resource contention or priority inversion ( §A). Observing that performance cost of Linux Audit is ultimately dependent on the number of log events generated, and that the performance impacts of commodity auditing frameworks can be optimized without affecting the forensic validity of the audit logs, e.g., through carefully reducing the number of events that need to be logged [43], [44], [45], [46], [47], [48], [37], [49], [50], we set out to tailor Linux Audit to RTS, carefully reducing event logging without impacting the forensic validity of the log. We present Ellipsis, a kernelbased log reduction framework that leverages the predictability of real-time tasksets' execution profiles. ...