| Commit message (Expand) | Author | Age |
| ... | |
| * | tenth pass: field-flow v2 on IDA boundaries with verified writers; TAs and AS... | user | 2026-09-29 |
| * | RETRACT the ninth pass: its positive control passed on garbage from a linear ... | user | 2026-09-29 |
| * | triage ASD's single field-flow site: bounded against the caller's capacity | user | 2026-09-29 |
| * | ninth pass: field-level detector passes both controls; TA negative is now con... | user | 2026-09-29 |
| * | eighth pass: build an inter-procedural detector, then decline to trust its ou... | user | 2026-09-29 |
| * | close the SYS_DRV underflow lead: the only caller bounds the length to 0x30 | user | 2026-09-29 |
| * | [ST] seventh pass: scan the code behind the TAs; one unresolved underflow sha... | user | 2026-09-29 |
| * | emulator fuzz harness: TAs really do load; first fault refuted on silicon | user | 2026-09-29 |
| * | sixth pass: build emulator fuzz harness for the twelve TAs; result deliberate... | user | 2026-09-29 |
| * | fifth pass: scan the whole corpus (~110 bodies) for all three shapes, build-d... | user | 2026-09-29 |
| * | fourth pass: scaled-index audit across the twelve TAs -- none host-controlled | user | 2026-09-29 |
| * | third pass: destination-side copy audit across the twelve TAs -- negative | user | 2026-09-29 |
| * | close every TA copy call site: bounded, except one latent callee-length shape | user | 2026-09-29 |
| * | [ST] latent XGMI stack overflow: callee-supplied length copied into a 68-byte... | user | 2026-09-29 |
| * | hunt a memory-corruption primitive in the twelve TAs: not found, and establis... | user | 2026-09-29 |
| * | assess the GCP defects' real capability: DoS alone, but an enabler for a key-... | user | 2026-09-29 |
| * | [ST] GCP version sweep: all three defects present in every harvested build | user | 2026-09-29 |
| * | [ST] audit the shared runtime clusters: negative, and the hypothesis is void | user | 2026-09-29 |
| * | [HW] retract the FE0A reachability claim: both new lookups are auth-gated | user | 2026-09-29 |
| * | [ST] sharpen the FE0F finding: two of the three lookups compare the never-wri... | user | 2026-09-29 |
| * | [ST] lifetime sweep across all twelve TAs: concentrated in GCP, and OPM prove... | user | 2026-09-29 |
| * | [ST] extend the FE0F finding: three liveness-free lookups in GCP, not one | user | 2026-09-29 |
| * | [ST] complete the ASD single-TA opcode sweep: uniformly guarded, no host path | user | 2026-09-29 |
| * | [ST] extend the opcode sweep to ASD: 0x403/0x404 closed, ASD now in the pipeline | user | 2026-09-29 |
| * | [ST] complete the SYS_DRV unique-opcode sweep: oc 0x109b benign, powerplay 0x... | user | 2026-09-29 |
| * | [ST] audit RAP's SYS_DRV opcodes; most of the ~23 unique opcodes are dead | user | 2026-09-29 |
| * | [ST] XGMI privileged-reach lead resolved negative: SYS_DRV stubs the whole block | user | 2026-09-29 |
| * | [ST] lead: XGMI cmd09/cmd0A pass host dwords into SYS_DRV 0x1061/0x1062 | user | 2026-09-29 |
| * | [ST] resolve the 0x60xx lead: emitter real, attacker-controlled value is not | user | 2026-09-29 |
| * | [ST] overturn the SYS_DRV 0x60xx 'no emitter' negative: POWERPLAY emits it | user | 2026-09-29 |
| * | fw: command-surface dumps for all twelve TAs, with honest handler attribution | user | 2026-09-29 |
| * | fw: regenerate the GCP command-surface dump after the arity pass | user | 2026-09-29 |
| * | fw: rebuild powerplay's corrupt database; verify all 632 functions decompile | user | 2026-09-29 |
| * | fw: recover helper arity by dataflow, validated first -- 632/632 functions no... | user | 2026-09-29 |
| * | fw: readable GCP command-surface dump + sharpen FE0F's in-database comment | user | 2026-09-29 |
| * | fw: restore AMD's own function names and apply prototypes/types across the TAs | user | 2026-09-29 |
| * | GCP acquire-gate bypass: measure the surface it opens, de-confound, audit it | user | 2026-09-29 |
| * | [HW] new finding: GCP crypto-session acquire gate bypassed by retrying | user | 2026-09-29 |
| * | ASD 0x202 path executes after the ASD-load fix, but the bytes are not a discl... | user | 2026-09-29 |
| * | GCP FE0F/FE0D: rule out UAF -- slot reuse rebinds the handle | user | 2026-09-29 |
| * | [HW] GCP FE0F/FE0D chain proven end-to-end: unauthorised lookup, destroy, repeat | user | 2026-09-29 |
| * | fw: document the annotation residue; all 632 functions now named or documented | user | 2026-09-29 |
| * | fw: verified final annotation coverage by querying every IDB directly (85%) | user | 2026-09-29 |
| * | psp_out: fold the twelve-TA round into the authoritative audit and disclosure... | user | 2026-09-29 |
| * | RESOLVE the MERIT cmd05 oracle: gated by an ASD-derived value; robustness def... | user | 2026-09-29 |
| * | fw: apply missed-function pass to all 5 affected TAs and re-run every detector | user | 2026-09-29 |
| * | LEAD: MERIT cmd05 compares with an unbounded host length; gated on hardware | user | 2026-09-29 |
| * | fw: IDA had missed functions -- coverage denominator and every sweep were inc... | user | 2026-09-29 |
| * | fw: characterise the unnamed remainder instead of inventing names; GCP to 113... | user | 2026-09-29 |
| * | fw: annotation coverage 65% -> 82%; POWERPLAY's fptr table found and named | user | 2026-09-29 |