Machine learning (ML) models are widely deployed on edge nodes, such as mobile phones and edge servers, to power a wide range of AI applications over the web. Ensuring the integrity of these edge models is paramount, as they are subject to corruption caused by software/hardware exceptions and malicious tampering, which may undermine model performance, incur economic losses, and pose health risks. Existing data integrity mechanisms designed for files stored on disks cannot properly verify the integrity of models running in GPUs or mitigate the new integrity threats against edge models. This paper proposes EdgeThemis, a novel mechanism for verifying the integrity of edge models through sentinel verification. To enable verifiability for a model M, EdgeThemis embeds a sentinel backdoor and a verification module into M. Then, a challenger can send verification requests to the edge node hosting M to verify its integrity. Next, the sentinel activates the verification module to generate a unique integrity proof tied to the identity of the edge node for verification. Finally, the challenger can verify the integrity proof to detect model corruption. Theoretical analysis proves that EdgeThemis can properly mitigate potential integrity threats against edge models. Experiments demonstrate that EdgeThemis achieves a verification accuracy of 100.00% across various models and different types of model corruption with robustness against replay attacks, theft attacks, and replacement attacks.

In an edge computing environment, data can be cached on edge servers to enable fast data services for users. These edge data are subject to corruption and must be verified to ensure their integrity. Meanwhile, they are also subject to partial content changes over time. Existing edge data integrity (EDI) schemes are designed to verify edge data as a whole. They fail to accommodate partially identical edge data and consequently suffer from low verification accuracy in many real-world applications. In the meantime, their reliability is subject to compromises caused by edge servers' Byzantine behaviors. This paper presents DynaEDI+, a novel decentralized EDI scheme capable of verifying the integrity of partially identical edge data. It introduces pairing trees, a new tree-based data digest structure, to enable subtree-based content comparison, allowing precise identification of version-matched data blocks. To reduce communication overhead, DynaEDI+ transmits only root nodes instead of entire trees for verification. It also implements a series of security mechanisms to safeguard the verification process against evasion attacks, replay attacks, and theft attacks from Byzantine edge servers. Theoretical analysis proves that DynaEDI+ can effectively defend against potential threats from Byzantine edge servers. Experimental results demonstrate that DynaEDI+ achieves high accuracy in edge environments with partially identical data and Byzantine edge servers, while reducing communication overhead by an order of magnitude compared to benchmark schemes.

Interrupt-driven programs have been widely used in safety-critical areas such as aerospace and embedded systems. However, uncertain interleaving execution of interrupt service routines (ISRs) usually causes concurrency bugs. Specifically, when one or more ISRs attempt to preempt a sequence of instructions which are expected to be atomic, a kind of concurrency bugs namely atomicity violation may occur, and it is challenging to find this kind of bugs precisely and efficiently. In this paper, we propose a static approach for detecting atomicity violations in interrupt-driven programs. First, the program model is constructed with interruption points being selected to determine the possibly influenced ISRs. After that, reachability computation is conducted to build up a whole abstract reachability tree, and a delayed ISR-triggering strategy is employed to reduce the state space. Meanwhile, unserializable interleaving patterns are recognized to achieve the goal of atomicity violation detection. The approach has been implemented as a configurable tool namely CPA4AV. Extensive experiments show that CPA4AV is much more precise than the relative tools available with little extra time overhead. In addition, more complex situations can be dealt with CPA4AV.

Abstract is a tool for verifying reachability properties of concurrent C programs. It moderates the trace-space explosion problem, aggravated by thread alternation, through utilizing the PC-DPOR and C-Intp techniques. The PC-DPOR technique constructs a constrained dependency graph to refine dependencies between transitions. With this basis, the inherent imprecision of the dependence over-approximation can be overcome. Thereby, many redundant equivalent traces are prevented from being explored. On the other hand, the C-Intp technique performs conditional interpolation to confine the reachable regions of states, so that infeasible conditional branches which occur more frequently in concurrent verification tasks could be pruned automatically. We have implemented the above techniques on top of the open-source program analysis framework .

Thread alternation aggravates the difficulty of concurrent program verification since the number of traces to be explored grows rapidly as the scale of a concurrent program increases. Partial-Order Reduction (POR) techniques alleviate the trace-space explosion problem by partitioning the traces into different equivalent classes. However, due to the coarse dependency approximation of transitions, there are still a large number of redundant traces explored throughout the verification. In this paper, a symbolic approach, namely Prioritized Constraint-Aided Dynamic Partial-Order Reduction (PC-DPOR), is proposed to reduce the redundant traces. Specifically, a constrained dependency graph is presented to refine dependencies between transitions, and the exploration of isolated transitions in the graph is prioritized to reduce redundant equivalent traces. Further, we utilize the generated constraints to dynamically detect whether the enabled transitions at the given reachable states are dependent, and thereby to overcome the inherent imprecision of the traditional dependence over-approximation. We have implemented the proposed approach as an extension of CPAchecker by utilizing BDDs as the representation of state sets. Experimental results show that our approach can effectively reduce the time and memory consumption for verifying concurrent programs. In particular, the number of explored states is reduced to 8.62% on average.