The Embedding Based Discovery Framework is introduced, which leverages continuous semantic embeddings to isolate latent collusive communities directly from their semantic profile, bypassing the adversarial limitations of explicit network analysis.
Abstract
The peer-review process, the bedrock of scientific advancement, is increasingly undermined by sophisticated collusion rings that systematically manipulate review outcomes to favor in-group members. While existing detection methods struggle to untangle obfuscated social ties in explicit co-authorship graphs, we introduce a new direction: Exclusion Based Anomaly Detection. Similar to the way VCG auctions work, we formally measure the marginal influence of suspected reviewer groups, exposing their signature even when explicit social graphs are hidden. To apply this at scale without prior knowledge of colluding groups, we introduce the Embedding Based Discovery Framework, which leverages continuous semantic embeddings to isolate latent collusive communities directly from their semantic profile, bypassing the adversarial limitations of explicit network analysis. Unlike traditional heuristic-based approaches, our framework functions as an automated auditor, requiring no prior knowledge of group membership. It achieves this by executing a decoupled search across independent diagnostic algorithms and combining their findings into distinct consensus formations, allowing organizers to dynamically balance detection precision and recall. Evaluating our technique with large-scale datasets (based on ICLR 2021) shows our method's capacity to identify both overt and subtle adversarial tactics with high sensitivity and strict Family-Wise Error Rate (FWER) control, effectively providing conference organizers with a scalable, robust, and privacy-preserving tool to secure the scientific integrity of academic publishing.
Sybil attackers are Blockchain actors that adopt the characteristics of regular users to exploit airdrops or influence governance. Current methods of Sybil actor detection include constructing graphs, which requires token transfers between examined wallets. Machine learning algorithms have been employed as well, but they treat the task as a closed-set classification problem, making them vulnerable to frequent changes in attack strategies or evasion tactics. We address the following questions: can compression-based similarity differentiate Sybil bots, organic users, and arbitrage bot wallets without direct financial links? What is the effect of high-signal contracts on the discovery of Sybils, and how robust are behavioral graphs under temporal drift and adversarial perturbations? Our approach synthesizes a symbolic Transaction Grammar from EVM (Ethereum Virtual Machine) traces, capturing separately transaction rhythm, execution structure, and functional intent. The high-signal contracts are filtered with our own protocol, called the Blind-Spot Protocol. Gzip-based NCD is used to construct a behavioral graph for Sybil discovery. We validate this framework against supervised machine learning baselines, a temporal split, and synthetic camouflage stress tests. Ultimately, we contribute a leakage-aware behavioral framework for Sybil candidate discovery. Its core NCD primitive requires no supervised training and can expand suspicious seed wallets without explicit funding links. We position the method as a training-free local discovery primitive for open-world blockchain audits, rather than as a formal open-set recognition system.
This work proposes an analyst-oriented clustering methodology based on Multi-Criteria Decision Analysis to surface operationally meaningful relationships among phishing messages that bypassed state-of-the-art protections and shows that the resulting Multi-Dimensional Clusters are cohesive, stable, and operationally meaningful.
Elyssa Boulila, Marc Dacier, Siva Prem et al.· 0 citations
Incentive programs are central to user acquisition in decentralized finance, but many reward systems rely on raw volume, transaction count, and wallet count, making them vulnerable to bots and sybil operations. We present ZAPs, a reward attribution framework that combines economic contribution scoring with adversarial robustness. A composite activity score uses protocol-specific percentile normalization to limit whale dominance while preserving differentiation among users. A two-layer weighting mechanism combines protocol share within sector and sector share within the ecosystem, which reduces the profitability of farming small protocols. We show that the maximum reward obtainable from any protocol is bounded by that protocol's global volume share. ZAPs also introduces a four-layer defense stack consisting of transaction-level integrity checks, a parallel anomaly ensemble, post-distribution behavioral memory, and graph-based sybil clustering. The anomaly ensemble combines a one-class reconstruction model with an isolation forest and applies graduated rather than binary penalties. On 1,073 labeled malicious wallets covering 124,638 transactions, the ensemble achieves 0.923 +/- 0.013 ROC-AUC, compared with 0.891 +/- 0.016 for the reconstruction model alone, when the isolation forest is trained on benign wallets. Training it on the pooled population reverses its polarity and removes the ensemble gain. Controlled simulations reduce adversarial reward capture by 30-90 percent while legitimate-user scenarios change by 1-8 percent. Live campaigns recorded a 56 percent reduction in sybil allocation, a 49 percent increase in quality-wallet participation, and a 50 percent reduction in sell pressure.
N. GirishG, Ashutosh Sahoo, Ajay Bhat et al.· 0 citations
Allegations that TikTok shadow bans political content shape what creators post, what advertisers fund, and how regulators act, yet they are hard to adjudicate because platforms do not disclose how content is ranked. We test the claim with a dense hourly panel of 556,946 follower-normalized views across 2,753 videos from 67 accounts curated into pro and anti sides of three contested topics (U.S. immigration enforcement, Trump coverage, and Israel/Palestine). On-topic videos are identified by a multi-step classifier, and stance is taken from each account's curated side. The conventional analysis appears to answer yes. Pooling the hourly snapshots, the topic-conditional reach gap reaches p<10^-140. Analyzed at the account level, the independent unit at which we sample and assign stance, the gap disappears. Every account-level reach contrast is null after correction (BH-FDR q near 0.9). We find no evidence of moderate-to-large reach suppression on any topic. The null is informative. Account-level confidence intervals and a power analysis rule out such effects. As a design check, the same framework detects a clear asymmetry on a different outcome. Oppositional content (anti-Trump, pro-Palestine) earns more engagement per view rather than less reach (Cliff's delta = -0.51 and -0.64; q<0.03). Higher engagement does not by itself rule out suppression, but shows the design can detect effects of this magnitude. The apparent reach gap is an artifact of two factors. The first is pseudoreplication, which counts tens of thousands of autocorrelated video-hours as independent observations; the second is confounding, since the side that looks suppressed is larger and, on Israel/Palestine, posts mostly in Arabic. In this corpus, what is taken for a shadow ban is better explained by a more engaged audience than by a suppressed one. We close with what a credible visibility audit requires.
An insider adversary manipulates the distance function to induce discriminatory clustering outcomes against demographic groups, exposing a gap in current fairness auditing practices and demonstrating the urgent need to expand the scope of the algorithmic audit to include distance-function verification.
Shahzad Ahmad, Stefan Rass, Enes Sovtic· Proceedings of the 23rd Inte...· 0 citations
Software supply-chain attacks increasingly exploit an identity gap where compromised maintainer accounts authorize malicious changes. This work evaluates patch-level authorship verification as a behavioral defense layer, showing that stylometric analysis can operate not only on full source files but also on patch-level commits. We fine-tune a cross-modal transformer on more than 20 years of Linux kernel commit history to embed code diffs and commit messages into a unified stylometric space, achieving ROC AUC of 0.93 for open-world authorship verification. We then use these representations in a streaming anomaly detector suited to continuous integration and deployment (CI/CD) settings. We validate the pipeline on two retrospective supply-chain incidents involving different patch characteristics: the 2021 PHP backdoor and the 2026 ForceMemo/GlassWorm campaign. Without retraining, the proposed detector surfaces both PHP forged commits within approximately 1% of the maintainer audit queue and ranks the 28 scoreable ForceMemo spoofs with a median per-repository review burden of 0.8%. These results indicate that cross-modal patch-level embeddings can support behavioral triage against author impersonation in real-world repositories.