Search

A common challenge towards the adaptability of Large Language Models (LLMs) is their ability to learn new languages over time without hampering the model's performance on languages in which the model is already proficient (usually English). Continual fine-tuning (CFT) is the process of sequentially fine-tuning an LLM to enable the model to adapt to downstream tasks with varying data distributions and time shifts. This paper focuses on the language adaptability of LLMs through CFT. We study a two-phase CFT process in which an English-only end-to-end fine-tuned LLM from Phase 1 (predominantly Task Ability) is sequentially fine-tuned on a multilingual dataset -- comprising task data in new languages -- in Phase 2 (predominantly Language Ability). We observe that the ``similarity'' of Phase 2 tasks with Phase 1 determines the LLM's adaptability. For similar phase-wise datasets, the LLM after Phase 2 does not show deterioration in task ability. In contrast, when the phase-wise datasets are not similar, the LLM's task ability deteriorates. We test our hypothesis on the open-source \mis\ and \llm\ models with multiple phase-wise dataset pairs. To address the deterioration, we analyze tailored variants of two CFT methods: layer freezing and generative replay. Our findings demonstrate their effectiveness in enhancing the language ability of LLMs while preserving task performance, in comparison to relevant baselines., Comment: 19 pages, 6 tables, 4 figures

Large language models (LLMs) have recently seen widespread adoption, in both academia and industry. As these models grow, they become valuable intellectual property (IP), reflecting enormous investments by their owners. Moreover, the high cost of cloud-based deployment has driven interest towards deployment to edge devices, yet this risks exposing valuable parameters to theft and unauthorized use. Current methods to protect models' IP on the edge have limitations in terms of practicality, loss in accuracy, or suitability to requirements. In this paper, we introduce a novel hybrid inference algorithm, named SLIP, designed to protect edge-deployed models from theft. SLIP is the first hybrid protocol that is both practical for real-world applications and provably secure, while having zero accuracy degradation and minimal impact on latency. It involves partitioning the model between two computing resources, one secure but expensive, and another cost-effective but vulnerable. This is achieved through matrix decomposition, ensuring that the secure resource retains a maximally sensitive portion of the model's IP while performing a minimal amount of computations, and vice versa for the vulnerable resource. Importantly, the protocol includes security guarantees that prevent attackers from exploiting the partition to infer the secured information. Finally, we present experimental results that show the robustness and effectiveness of our method, positioning it as a compelling solution for protecting LLMs.

Reviews and ratings by users form a central component in several widely used products today (e.g., product reviews, ratings of online content, etc.), but today's platforms for managing such reviews are ad-hoc and vulnerable to various forms of tampering and hijack by fake reviews either by bots or motivated paid workers. We define a new metric called 'hijack-resistance' for such review platforms, and then present TrustRate, an end-to-end decentralized, hijack-resistant platform for authentic, anonymous, tamper-proof reviews. With a prototype implementation and evaluation at the scale of thousands of nodes, we demonstrate the efficacy and performance of our platform, towards a new paradigm for building products based on trusted reviews by end users without having to trust a single organization that manages the reviews., Comment: 23 pages. Poster at The 24th Privacy Enhancing Technologies Symposium, 2024, Bristol, United Kingdom

We analyze the number of queries that a whitebox adversary needs to make to a private learner in order to reconstruct its training data. For $(\epsilon, \delta)$ DP learners with training data drawn from any arbitrary compact metric space, we provide the \emph{first known lower bounds on the adversary's query complexity} as a function of the learner's privacy parameters. \emph{Our results are minimax optimal for every $\epsilon \geq 0, \delta \in [0, 1]$, covering both $\epsilon$-DP and $(0, \delta)$ DP as corollaries}. Beyond this, we obtain query complexity lower bounds for $(\alpha, \epsilon)$ R\'enyi DP learners that are valid for any $\alpha > 1, \epsilon \geq 0$. Finally, we analyze data reconstruction attacks on locally compact metric spaces via the framework of Metric DP, a generalization of DP that accounts for the underlying metric structure of the data. In this setting, we provide the first known analysis of data reconstruction in unbounded, high dimensional spaces and obtain query complexity lower bounds that are nearly tight modulo logarithmic factors., Comment: Updated proof of Thm 10, fixed typos

The use of blockchain in regulatory ecosystems is a promising approach to address challenges of compliance among mutually untrusted entities. In this work, we consider applications of blockchain technologies in telecom regulations. In particular, we address growing concerns around Unsolicited Commercial Communication (UCC aka. spam) sent through text messages (SMS) and phone calls in India. Despite several regulatory measures taken to curb the menace of spam it continues to be a nuisance to subscribers while posing challenges to telecom operators and regulators alike. In this paper, we present a consortium blockchain based architecture to address the problem of UCC in India. Our solution improves subscriber experiences, improves the efficiency of regulatory processes while also positively impacting all stakeholders in the telecom ecosystem. Unlike previous approaches to the problem of UCC, which are all ex-post, our approach to adherence to the regulations is ex-ante. The proposal described in this paper is a primary contributor to the revision of regulations concerning UCC and spam by the Telecom Regulatory Authority of India (TRAI). The new regulations published in July 2018 were first of a kind in the world and amended the 2010 Telecom Commercial Communication Customer Preference Regulation (TCCCPR), through mandating the use of a blockchain/distributed ledgers in addressing the UCC problem. In this paper, we provide a holistic account of of the projects' evolution from (1) its design and strategy, to (2) regulatory and policy action, (3) country wide implementation and deployment, and (4) evaluation and impact of the work., Comment: 20 pages, 6 figures, 1 table

We introduce Blockene, a blockchain that reduces resource usage at member nodes by orders of magnitude, requiring only a smartphone to participate in block validation and consensus. Despite being lightweight, Blockene provides a high throughput of transactions and scales to a large number of participants. Blockene consumes negligible battery and data in smartphones, enabling millions of users to participate in the blockchain without incentives, to secure transactions with their collective honesty. Blockene achieves these properties with a novel split-trust design based on delegating storage and gossip to untrusted nodes. We show, with a prototype implementation, that Blockene provides throughput of 1045 transactions/sec, and runs with very low resource usage on smartphones, pointing to a new paradigm for building secure, decentralized applications., Comment: A version of this paper (without the appendix) will appear in OSDI 2020

Let $\mathcal{F}_{n}^*$ be the set of Boolean functions depending on all $n$ variables. We prove that for any $f\in \mathcal{F}_{n}^*$, $f|_{x_i=0}$ or $f|_{x_i=1}$ depends on the remaining $n-1$ variables, for some variable $x_i$. This existent result suggests a possible way to deal with general Boolean functions via its subfunctions of some restrictions. As an application, we consider the degree lower bound of representing polynomials over finite rings. Let $f\in \mathcal{F}_{n}^*$ and denote the exact representing degree over the ring $\mathbb{Z}_m$ (with the integer $m>2$) as $d_m(f)$. Let $m=\Pi_{i=1}^{r}p_i^{e_i}$, where $p_i$'s are distinct primes, and $r$ and $e_i$'s are positive integers. If $f$ is symmetric, then $m\cdot d_{p_1^{e_1}}(f)... d_{p_r^{e_r}}(f) > n$. If $f$ is non-symmetric, by the second moment method we prove almost always $m\cdot d_{p_1^{e_1}}(f)... d_{p_r^{e_r}}(f) > \lg{n}-1$. In particular, as $m=pq$ where $p$ and $q$ are arbitrary distinct primes, we have $d_p(f)d_q(f)=\Omega(n)$ for symmetric $f$ and $d_p(f)d_q(f)=\Omega(\lg{n}-1)$ almost always for non-symmetric $f$. Hence any $n$-variate symmetric Boolean function can have exact representing degree $o(\sqrt{n})$ in at most one finite field, and for non-symmetric functions, with $o(\sqrt{\lg{n}})$-degree in at most one finite field., Comment: 5 pages, ISIT 2015. Simplified proof