Abstract:Programs increasingly rely on randomization in applications such as cryptography and machine learning. Analyzing randomized programs has been a fruitful research direction, but there is a gap when programs also exploit nondeterminism (for concurrency, efficiency, or algorithmic design). In this paper, we introduce Demonic Outcome Logic for reasoning about programs that exploit both randomization and nondeterminism. The logic includes several novel features, such as reasoning about multiple executions in tandem and manipulating pre- and postconditions using familiar equational laws -- including the distributive law of probabilistic choices over nondeterministic ones. We also give rules for loops that both establish termination and quantify the distribution of final outcomes from a single premise. We illustrate the reasoning capabilities of Demonic Outcome Logic through several case studies, including the Monty Hall problem, an adversarial protocol for simulating fair coins, and a heuristic based probabilistic SAT solver.

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve This paper attempts to address the analysis problem when utilizing both randomness and nondeterminism (random nondeterminism) in programs. Specifically, as the application of randomness in fields such as cryptography and machine learning becomes increasingly widespread, analyzing random programs has become a productive research direction. However, there is currently a gap when programs also utilize nondeterminism (for concurrency, efficiency, or algorithm design). Existing verification techniques can handle programs with both randomness and nondeterminism, but there is currently no logic that can describe and reason about the multiple possible execution paths of these programs and their result distributions. ### Main Contributions 1. **From Equations to Propositions**: - Proposed a new assertion language that allows logical implication to reflect equation laws. - Designed an assertion language that makes these laws hold. 2. **Demonic Outcome Logic**: - Introduced a new program logic for reasoning about programs with both randomness and nondeterminism. - Created simple and convenient reasoning rules through demonic postconditions (i.e., postconditions apply to every nondeterministic possibility). - Compared to Weakest Pre-Expectation calculus, Demonic Outcome Logic can reason about multiple execution paths simultaneously, specifying the distribution of results rather than just quantitative properties. 3. **Loops and Termination**: - Proposed reasoning rules for loops that can prove termination and simultaneously specify the result distribution upon termination. - These rules are simpler and require fewer preconditions than expectation-based reasoning methods. 4. **Case Studies**: - Demonstrated the application of the logic through three case studies, including the Monty Hall problem, an adversarial protocol for simulating a fair coin, and a heuristic-based probabilistic SAT solver. - Proved that these programs can terminate under certain conditions and that the result distributions meet expectations. ### Abstract As the importance of randomness in sensitive software fields (such as cryptography and machine learning) increases, analyzing random programs has become an important research direction. However, when these programs also utilize nondeterminism, the analysis becomes more complex. This paper introduces a new program logic—Demonic Outcome Logic—for reasoning about programs with both randomness and nondeterminism. This logic includes several novel features, such as reasoning about multiple execution paths simultaneously, familiar equation laws for manipulating preconditions and postconditions (including the distributive law of probabilistic choice over nondeterministic choice), and rules for loops that can prove termination and quantify the final result distribution. Through several case studies, the reasoning capabilities of Demonic Outcome Logic are demonstrated.

A Demonic Outcome Logic for Randomized Nondeterminism

Probabilistic Concurrent Reasoning in Outcome Logic: Independence, Conditioning, and Invariants

Non Deterministic Logic Programs

Unifying Probability with Nondeterminism

The algebra of non-deterministic programs: demonic operators, orders and axioms

Outcome Logic: A Unifying Foundation for Correctness and Incorrectness Reasoning

Total Outcome Logic: Proving Termination and Nontermination in Programs with Branching

Verification of Nondeterministic Quantum Programs

Outcome Separation Logic: Local Reasoning for Correctness and Incorrectness with Computational Effects

Improved Parallel Derandomization via Finite Automata with Applications

Inductive Coherence

Demonic variance and a non-determinism score for Markov decision processes

Guaranteed Bounds on Posterior Distributions of Discrete Probabilistic Programs with Loops

Non-Determinism and the Lawlessness of Machine Learning Code

Does a Program Yield the Right Distribution? Verifying Probabilistic Programs via Generating Functions

DynamiTe: Dynamic Termination and Non-termination Proofs

When black box algorithms are (not) appropriate: a principled prediction-problem ontology

On Separation Logic, Computational Independence, and Pseudorandomness (Extended Version)

Does a Program Yield the Right Distribution?

Bird's-eye view on Noise-Based Logic

Combining Classical and Probabilistic Independence Reasoning to Verify the Security of Oblivious Algorithms (Extended Version)