Adversarial attacks pose severe threats to the integrity of deep neural networks (DNNs), especially in resource-constrained systems where traditional defenses are computationally expensive. While existing defenses in the black-box setting utilize hardware characteristics of adversarial attacks (like Hardware Performance Counter or HPC measurements), these defenses often involve repeating execution of multiple target model inferences to detect the attacks.In this work, we put forth a differing perspective: while detection strategies involving multiple target model inferences appear to be successful in isolation, they have unacceptable and inhibitory requirements. Precisely, we argue that these works require cleaning the micro-architectural state of hardware like the cache and the branch predictor after each inference. This in turn leads to performance degradation of not only the adversarial attack detector, but also of the overall system at large.In this work, we put forth a novel and lightweight detection strategy, MIRAGE, using HPCs that does not require cleaning the micro-architectural state of caches or branch predictors. We train a convolutional neural network (CNN) on these signals to classify inputs as benign or adversarial in a single shot, making our approach practical for online systems, while allowing full use of hardware optimizations for performance uplifts. Experiments on CIFAR-10 and MNIST datasets reveal that our methodology not only detects adversarial samples effectively with greater than 96% accuracy, but also imposes a minimal timing overhead of 60 ms and maintains high throughput. This makes our solution well-suited for embedded and edge-AI scenarios.

Searchable Symmetric Encryption (SSE) supports efficient keyword searches over encrypted outsourced document collections while minimizing information leakage. All practically efficient SSE schemes supporting conjunctive queries rely crucially on quantum-broken cryptographic assumptions (such as discrete-log hard groups) to achieve compact storage and fast query processing. On the other hand, quantum-safe SSE schemes based on purely symmetric-key cryptoprimitives either do not support conjunctive searches, or are practically inefficient. In particular, there exists no quantum-safe yet practically efficient conjunctive SSE scheme from lattice-based hardness assumptions.We solve this open question by proposing Oblivious Post-Quantum Secure Cross Tags (OQXT) – the first lattice-based practically efficient and highly scalable conjunctive SSE scheme. The technical centerpiece of OQXT is a novel oblivious cross-tag generation protocol with provable security guarantees derived from lattice-based hardness assumptions. We prove the post-quantum simulation security of OQXT with respect to a rigorously defined and thoroughly analyzed leakage profile. We then present a prototype implementation of OQXT and experimentally validate its practical efficiency and scalability over extremely large real-world databases. Our experiments show that OQXT has competitive end-to-end search latency when compared with the best (quantum-broken) conjunctive SSE schemes.

Searchable Symmetric Encryption (SSE) supports efficient yet secure query processing over outsourced symmetrically encrypted databases without the need for decryption. A longstanding open question has been the following: can we design a fast, scalable, linear storage and low-leakage SSE scheme that efficiently supports arbitrary Boolean queries over encrypted databases? In this paper, we present the design, analysis and prototype implementation of the first SSE scheme that efficiently supports conjunctive, disjunctive and more general Boolean queries (in both the conjunctive and disjunctive normal forms) while scaling smoothly to extremely large encrypted databases, and while incurring linear storage overheads and supporting extremely fast query processing in practice. We quantify the leakage of our proposal via a rigorous cryptographic analysis and argue that it achieves security against a well-known class of leakage-abuse and volume analysis attacks. Finally, we demonstrate the storage-efficiency and scalability of our proposed scheme by presenting experimental results of a prototype implementation of our scheme over large real-world databases.