As Large Language Models (LLMs) become increasingly widespread, understanding how specific training data shapes their outputs is crucial for transparency, accountability, privacy, and fairness. To explore how LLMs leverage and replicate their training data, we introduce a systematic approach centered on analyzing low-perplexity sequences - high-probability text spans generated by the model. Our pipeline reliably extracts such long sequences across diverse topics while avoiding degeneration, then traces them back to their sources in the training data. Surprisingly, we find that a substantial portion of these low-perplexity spans cannot be mapped to the corpus. For those that do match, we quantify the distribution of occurrences across source documents, highlighting the scope and nature of verbatim recall and paving a way toward better understanding of how LLMs training data impacts their behavior.

The study by Gohr et.al at CRYPTO 2019 and sunsequent related works have shown that neural networks can uncover previously unused features, offering novel insights into cryptanalysis. Motivated by these findings, we employ neural networks to learn features specifically related to integral properties and integrate the corresponding insights into optimized search frameworks. These findings validate the framework of using neural networks for feature exploration, providing researchers with novel insights that advance established cryptanalysis methods. Neural networks have inspired the development of more precise integral search models. By comparing the integral distinguishers obtained via neural networks with those identified by classical methods, we observe that existing automated search models often fail to find optimal distinguishers. To address this issue, we develop a meet in the middle search framework that balances model accuracy and computational efficiency. As a result, we reduce the number of active plaintext bits required for an 11 rounds integral distinguisher on SKINNY64/64, and further identify a 12 rounds key dependent integral distinguisher achieving one additional round over the previous best-known result. The integral distinguishers discovered by neural networks enable key recovery attacks on more rounds. We identify a 7 rounds key independent integral distinguisher from neural networks with even only one active plaintext cell, which is based on linear combinations of bits. This distinguisher enables a 15 rounds key recovery attack on SKINNYn/n, improving upon the previous record by one round. Additionally, we discover an 8 rounds key dependent integral distinguisher using neural network that further reduces the time complexity of key recovery attacks against SKINNY.

The fields of machine learning (ML) and cryptanalysis share an interestingly common objective of creating a function, based on a given set of inputs and outputs. However, the approaches and methods in doing so vary vastly between the two fields. In this paper, we explore integrating the knowledge from the ML domain to provide empirical evaluations of cryptosystems. Particularly, we utilize information theoretic metrics to perform ML-based distribution estimation. We propose two novel applications of ML algorithms that can be applied in a known plaintext setting to perform cryptanalysis on any cryptosystem. We use mutual information neural estimation to calculate a cryptosystem's mutual information leakage, and a binary cross entropy classification to model an indistinguishability under chosen plaintext attack (CPA). These algorithms can be readily applied in an audit setting to evaluate the robustness of a cryptosystem and the results can provide a useful empirical bound. We evaluate the efficacy of our methodologies by empirically analyzing several encryption schemes. Furthermore, we extend the analysis to novel network coding-based cryptosystems and provide other use cases for our algorithms. We show that our classification model correctly identifies the encryption schemes that are not IND-CPA secure, such as DES, RSA, and AES ECB, with high accuracy. It also identifies the faults in CPA-secure cryptosystems with faulty parameters, such a reduced counter version of AES-CTR. We also conclude that with our algorithms, in most cases a smaller-sized neural network using less computing power can identify vulnerabilities in cryptosystems, providing a quick check of the sanity of the cryptosystem and help to decide whether to spend more resources to deploy larger networks that are able to break the cryptosystem.

In this research, we introduce MIND-Crypt, a novel attack framework that uses deep learning (DL) and transfer learning (TL) to challenge the indistinguishability of block ciphers, specifically SPECK32/64 encryption algorithm in CBC mode (Cipher Block Chaining) against Known Plaintext Attacks (KPA). Our methodology includes training a DL model with ciphertexts of two messages encrypted using the same key. The selected messages have the same byte-length and differ by only one bit at the binary level. This DL model employs a residual network architecture. For the TL, we use the trained DL model as a feature extractor, and these features are then used to train a shallow machine learning, such as XGBoost. This dual strategy aims to distinguish ciphertexts of two encrypted messages, addressing traditional cryptanalysis challenges. Our findings demonstrate that the DL model achieves an accuracy of approximately 99% under consistent cryptographic conditions (Same Key or Rounds) with the SPECK32/64 cipher. However, performance degrades to random guessing levels (50%) when tested with ciphertext generated from different keys or different encryption rounds of SPECK32/64. To enhance the results, the DL model requires retraining with different keys or encryption rounds using larger datasets (10^7 samples). To overcome this limitation, we implement TL, achieving an accuracy of about 53% with just 10,000 samples, which is better than random guessing. Further training with 580,000 samples increases accuracy to nearly 99%, showing a substantial reduction in data requirements by over 94%. This shows that an attacker can utilize machine learning models to break indistinguishability by accessing pairs of plaintexts and their corresponding ciphertexts encrypted with the same key, without directly interacting with the communicating parties.

The case of the Zodiac Killer is one of the most widely known unsolved serial killer cases in history. The unidentified killer murdered five known victims and terrorized the state of California. He also communicated extensively with the press and law enforcement. Besides his murders, Zodiac was known for his use of ciphers. The first Zodiac cipher was solved within a week of its publication, while the second cipher was solved by the authors after 51 years, when it was discovered to be a transposition and homophonic substitution cipher with unusual qualities. In this paper, we detail the historical significance of this cipher and the numerous efforts which culminated in its solution.

In the not so unlikely scenario of total compromise of computers accessible to a group of users, they might be tempted to resort to human-computable paper-and-pencil cryptographic methods aided by a classic Tabula Recta, which helps to perform addition and subtraction directly with letters. But do these classic algorithms, or some new ones using the same simple tools, have any chance against computer-aided cryptanalysis? In this paper I discuss how some human-computable algorithms can indeed afford sufficient security in this situation, drawing conclusions from computer-based statistical analysis. Three kinds of algorithms are discussed: those that concentrate entropy from shared text sources, stream ciphers based on arithmetic of non-binary spaces, and hash-like algorithms that may be used to generate a password from a challenge text.

Over-parameterized neural language models (LMs) can memorize and recite long sequences of training data. While such memorization is normally associated with undesired properties such as overfitting and information leaking, our work casts memorization as an unexplored capability of LMs. We propose the first symmetric encryption algorithm with autoregressive language models (SELM). We show that autoregressive LMs can encode arbitrary data into a compact real-valued vector (i.e., encryption) and then losslessly decode the vector to the original message (i.e., decryption) via random subspace optimization and greedy decoding. While SELM is not amenable to conventional cryptanalysis, we investigate its security through a novel empirical variant of the classic IND-CPA (indistinguishability under chosen-plaintext attack) game and show promising results on security. Our code and datasets are available at https://github.com/OSU-NLP-Group/SELM.

Will artificial intelligence become clever enough to upend computer security? AI is already surprising the world of art by producing masterpieces in any style on demand. If AIs can act like a bard while delivering the comprehensive power of the best search engines, why can't they shatter security protocols, too? The answers are complex, rapidly evolving, and still murky. AI makes some parts of defending computers against attack easier.