Text of github.com › jegly as Styx crawled it on Nov 19, 2026. The page may have changed since. Only its words are kept, without pictures or layout, and only the first part of a long page.

Open the live page

jegly (Jesse Li-Yates) · GitHub

Skip to content You signed in with another tab or window. Reload to refresh your session. You signed out in another tab or window. Reload to refresh your session. You switched accounts on another tab or window. Reload to refresh your session. Dismiss alert jegly Follow More Overview Repositories Projects Packages Stars jegly Follow Jesse Li-Yates jegly Follow Sponsor I make Android & Linux apps with security and privacy in mind.

114 followers · 0 following Sydney,Australia https://www.jegly.xyz Block or report jegly Report abuse Contact GitHub support about this user’s behavior. Learn more about reporting abuse . Report abuse More Overview Repositories Projects Packages Stars jegly / README .md I've carried out independent security research for a number of vendors and OEMs, identifying vulnerabilities that have resulted in several CVEs. I've also worked directly with various vendors and manufacturers on security research and vulnerability disclosures.

FYI I'm not against Google.. One of my own projects, Box, started as a fork of their AI Edge Gallery, and I've worked on it most days for the past year. Google has pulled features from my fork back into their project, so it has gone both ways, This is specifically about what happened when I reported real, working security bugs through their own process over a period of 12 months: they used the work, and they didn't pay for it..

I also maintain Iris which is a hardened Chromium fork focused on Modern day security and privacy, so as you can see i work on several google related projects. For most of the last year this is what I've done. I reverse engineered the Samsung radio stack (under embargo), the IMS and RTP media code, the Trusty trusted apps, the MediaTek M90 modem RIL, and the Android telephony framework on Google's current phones, and many many more things.

I filed ninety five reports through the Android and Google Devices Vulnerability Reward Program. 95% had a working proof of concept. Total paid, for all of it: $250USD. One payment. This isn't one unlucky ticket. The same thing keeps happening three different ways. They fixed my bugs after closing them. I pulled each closed report's binary out of Google's current factory image and diffed the exact function the report named.

Eight of them had been fixed in shipped code: a bounds check, a permission, a validation step added exactly where I said it was missing. Every one of those eight was first closed as a "duplicate" or "not a vulnerability." None was paid. None was credited. Two more have since been fixed on the QPR3 beta. They closed bugs with reasons the code contradicts. More than a dozen others were closed duplicate, infeasible, or "can't reproduce," and the exact defect is still sitting unchanged in the current build.

Nobody owes a reward for an unfixed bug, and I'm not claiming one. The point is narrower: the stated reason for closing them does not survive a look at the actual binary. They promised a reward in writing, in January, and still haven't paid it. Google confirmed a valid bug, raised the severity, raised the reward, in writing. Three months later they reversed all three with no new evidence. Then they said they would "stand by our word" and pay it anyway as a $250 "discretionary payment exception." That was January.

Ten months of "processing / soon / no date" later, it still isn't paid,but was submitted prior to December 2025. One line from that thread I still can't parse. They told me, in writing, that they are paying the Safety Center reward "even though our final technical assessment rated it as a duplicate of a report with no security impact." If it has no security impact there is no basis to pay it. If it is worth paying, "no security impact" is false.

Both are in the ticket. Where this stands right now $250, Safety Center report (Moderate). Approved in writing in January, reconfirmed in writing several times since. Unpaid ten months later. Ticket closed... Two High severity reports, still "Assigned." Both formally rated High by Google. Rewards undetermined, CVEs awaiting to be assigned. One of the two has already been fixed three times in shipped code: first in the QPR3 beta (DP11.261002.005) in October, then in the public November stable (CP4A.261105.040.D1, rolled out to all Pixels on 2026-11-10), and again in the latest QPR3 beta (DP11.261030.003, released 12 November 2026), and the ticket is STILL "Assigned" with the reward field blank.. ?!?!?

One Moderate report, still "Assigned." Rated Moderate on 28 October. Reward undetermined. Ten findings fixed after closure. Closed as duplicate, infeasible, or not-a-vulnerability, then fixed in shipped code. No reward, no credit.. !? Total actually received across all of it: $250.. Everything in the tables below is a closed report, with Google's own words quoted from the tracker. The three still "Assigned" reports and anything under active coordinated disclosure are kept out of the technical detail on purpose.

How I checked Same method for every closed report in the tables: Pull the exact binary, library, or jar the report named, out of Google's current factory image (rango and yogi Pixel builds, September and October 2026). Extract it with the right tool for the container: debugfs for ext4, fsck.erofs for EROFS, a second pass through apex_payload.img for APEX modules. Find the exact function the report cited. Symbol resolution where symbols survive, byte pattern matching against the report's own quoted disassembly where they don't.

Diff the current shipped code against what the report called vulnerable, using the NDK llvm-objdump . (The system binutils has no AArch64 backend and returns nothing without an error, which is worth knowing before you trust an empty diff.) Verdicts are PATCHED (a check was added that wasn't there when I reported it), STILL PRESENT (code unchanged), N/A (reported against this exact build, or Google called it intended design), and CANT-DETERMINE (binary absent or build mismatch).

I didn't use Google's bulletin or supplemental patch list for this. A silent fix with no bulletin entry and no credit doesn't appear there by definition. You have to diff the shipped code, which is what this is.. The reward they promised in writing and still haven't paid On ticket 467575499 / 470814942 , on 23 January 2026 , Google wrote that they had "confirmed this is a valid bug," were "increasing severity," "increasing reward," and would "issue a payment for the new amount once it has been processed." On 24 April , with no new analysis, they reversed all three.

The same bug now "does not qualify for a reward or severity increase," called a "Final Assessment." In June they wrote that they wanted to "stand by our word regarding the reward." On 10 July they gave the numbers: "$250 for Safety Center, $250 for the mic bug, $500USD total," and said they were "honoring our initial promise," while still calling the Safety Center half a duplicate with no security impact. The mic half was paid.

The Safety Center half went into a "discretionary payment exception" with, in their words, no "exact credit date." That was January. The most recent reply, in October, says the same thing: still approved, no date. 23 Jan 2026: "confirmed this is a valid bug," severity increased, reward increased, "we will issue a payment for the new amount once it has been processed." 24 Apr 2026: all three reversed with no new analysis.

The same report now "does not qualify for a reward or a severity increase." Labelled a "Final Assessment." 9 Jul 2026: paying it anyway "even though our final technical assessment rated it as a duplicate of a report with no security impact." 10 Jul 2026: the numbers, in writing. $250 + $250, $500 total. The Safety Center half put into a "discretionary payment exception" with no "exact credit date." Sidebar: ASR Payment 250.

31 Oct 2026: still no date, no Bugcrowd reference. "your $250 reward is fully approved." And: "We also respect your right to publish a public write-u

The rest of this page is on the live site.